HAL Id: tel-01668439 https://tel.archives-ouvertes.fr/tel-01668439 Submitted on 20 Dec 2017 HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés. Development of time-dependent characterisation factors for life cycle impact assessment of road traffc noise on human health. Rodolphe Meyer To cite this version: Rodolphe Meyer. Development of time-dependent characterisation factors for life cycle impact as- sessment of road traffc noise on human health.. Human health and pathology. Université de Cergy Pontoise, 2017. English. NNT : 2017CERG0879. tel-01668439
151
Embed
Development of time-dependent characterisation factors for ...
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
HAL Id: tel-01668439https://tel.archives-ouvertes.fr/tel-01668439
Submitted on 20 Dec 2017
HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, estdestinée au dépôt et à la diffusion de documentsscientifiques de niveau recherche, publiés ou non,émanant des établissements d’enseignement et derecherche français ou étrangers, des laboratoirespublics ou privés.
Development of time-dependent characterisation factorsfor life cycle impact assessment of road traffic noise on
human health.Rodolphe Meyer
To cite this version:Rodolphe Meyer. Development of time-dependent characterisation factors for life cycle impact as-sessment of road traffic noise on human health.. Human health and pathology. Université de CergyPontoise, 2017. English. �NNT : 2017CERG0879�. �tel-01668439�
factors for life cycle impact assessment of road traffic
noise on human health
Présentée par Rodolphe Meyer
Pour obtenir le grade de Docteur de l’Université de Cergy-Pontoise en sciences et technologies de l'information et de la communication
Spécialité : Sciences de l'environnement
Soutenue le 10 novembre 2017
Membres du jury :
Catherine Lavandier, Directeur de thèse
Rosario Vidal, Rapporteur
Manuele Margni, Rapporteur
Enrico Benetto, Examinateur
Dick Botteldooren, Examinateur
Maarten Messagie, Examinateur
Benoit Gauvreau, Examinateur
Frédéric Mauny, Examinateur et président du jury
Bruno Vincent, Invité
i
“Soon silence will have passed into legend. Man has turned his back on silence. Day after day he invents machines and devices that increase noise and distract humanity from the essence of life, contemplation,
meditation...tooting, howling, screeching, booming, crashing, whistling, grinding, and trilling bolster his ego. His anxiety subsides. His inhuman void spreads monstrously like a gray vegetation.”
― Jean Arp
Even if my PhD thesis only addresses environmental noise impact on human health, I want to put it in a wider context, the context of our era, our epoch: the Anthropocene. This term is not yet officially accepted by the geological community as a geological period. However, to anybody acknowledging the state of our common spaceship Earth, this word has a clear meaning. It is the epoch where the consequences of human activities are visible everywhere. Human activities have halved the world’s animal population. The climate, and thus the physical condition of life on our planet, is changing at an unprecedented speed. Oceans are acidifying to the point of putting in danger the life it contains, especially the unique coral reef ecosystems. New minerals are created and major substance cycles, such as nitrogen and phosphorus, are visibly modified. Wherever we look, the impacts of human activities dominate natural variations. The lights of the city replace the fireflies. Concrete spreads over the fields of past generations covering productive soils and pushing agricultural activities further away, leaving less and less space for wildlife. Noise is not an exception in the great picture of the Anthropocene. Soundscape ecologists have found human noise everywhere; anthropophony is taking over biophony and geophony. Chain saws and car alarms are more and more common in the songs of the lyrebirds.
The environmental conditions are changing for the worse. We are not only damaging ecosystems, but we are also damaging the ability of our planet to support our species. The Anthropocene is the epoch where a single species is crossing the planetary boundaries. In a few decades, humanity has passed from a world where the limits were so far that they were imperceptible to a world where the limits are so close that they define the possible space, narrower every day, in which our societies can move if they want to keep what is perceived as civilized state. The Anthropocene is about a world with physical limits, and for humanity, it is something new.
Despite scientific facts, neither societies, nor populations, nor economic thinking has yet to successfully grasp the multiple implications of this drastic change of perspective. Modern societies rely heavily on non-renewable resources and produce more pollution than the environment is able to absorb. For a large part, today’s thinkers are still using old models to think about the future or are taking refuge in the faith of an hypothetic technologic rescue. If scientists and facts are not able to weigh in the way we think and build the future, my generation will quickly face the unthought of, or worse, the unthinkable. The last generations, despite the efforts of many, did not drastically change the way humanity considers itself and its environment. We will not have the same luxury. The decades to come may be the most important in the evolutionary process of our civilization.
ii
Table of Contents 1 Introduction .................................................................................................................................................... 3
1.1 Life cycle assessment (LCA) ................................................................................................................ 3
2.2 General consideration about noise and noise indicators ........................................................................ 9
2.3 Health impairments ............................................................................................................................. 11
4 Noise prediction model ................................................................................................................................. 18
4.1 Required data ...................................................................................................................................... 18
4.2 From noise emission to population exposure ...................................................................................... 18
4.2.1 Noise emission model ..................................................................................................................... 19
4.2.2 Noise propagation model ................................................................................................................ 20
4.2.3 Population exposure ........................................................................................................................ 21
4.3 Maximum search radius ...................................................................................................................... 22
6.4.6 Influence of other sources of uncertainty for an average CF, CFavg,day,IRIS ...................................... 71
iv
6.5 Comparison to previous literature ....................................................................................................... 74
6.5.1 Environmental noise impacts on human health .............................................................................. 74
6.5.2 Distance-based CFs in the LCA literature ...................................................................................... 75
6.5.3 The specific case of Cucurachi ....................................................................................................... 77
6.6 Connecting to the inventory ................................................................................................................ 78
6.6.1 Reference vehicle used ................................................................................................................... 78
6.6.2 New elementary flows .................................................................................................................... 78
6.6.3 Calculating the output of the processes........................................................................................... 79
6.7 Practical use of the CFs ....................................................................................................................... 81
6.7.1 In existing road transportation processes ........................................................................................ 81
6.7.2 In the foreground ............................................................................................................................ 82
7.4 Using the IRIS division of the territory as basis for the calculation .................................................... 92
7.5 Pertinence of spatialized CFs for the environmental noise question ................................................... 93
7.6 Distance-based approach and vehicle categories ................................................................................ 94
7.7 Use of marginal and average CFs ....................................................................................................... 95
7.8 Temporal differentiation of environmental noise impact on human health......................................... 96
7.9 Uncertainty of environmental noise impact ........................................................................................ 97
8 Conclusion and outlook ................................................................................................................................ 98
10 Annex I .................................................................................................................................................. 105
11 Annex II ................................................................................................................................................. 120
Table 11: Mean number of joules per vkm calculated from the weighted mean of the marginal CFs. ________ 65
Table 12: Comparing average and marginal CFs at the midpoint and endpoint level. _____________________ 65
Table 13: First order and total-effect Sobol indices of the considered uncertainties. ______________________ 70
Table 14: First order and total-effect Sobol indices of the considered uncertainties. ______________________ 74
Table 15: Comparison of the results of Müller-Wenk (2002) and the present work. ______________________ 75
Table 16: CFs calculated with the approach proposed by Cucurachi and CFs ____________________________ 78
Table 17: New elementary flows for distance-based CFs. ___________________________________________ 79
Table 18: Mean speed in km/h of LVs and HGVs for the weighted mean of marginal CFs. _________________ 81
Table 19: Human health impacts for one vkm. ___________________________________________________ 81
Table 20: Uncertainty distributions for the different parameters _____________________________________ 83
Table 21: Results of the uncertainty analysis. ____________________________________________________ 84
Table 22: Total-effect Sobol index for the five most important variables. ______________________________ 84
vii
Recurring acronyms
CF
DALY*
DW
EF
GIS
HAP*
HGV
HSDP*
IRIS
LCA
LCI
LCIA
LV
NEPM
pkm
SM
tkm
vkm
WHO
Characterisation factor
Disability-adjusted life year
Disability weight
Elementary flow
Geographical information system
Highly annoyed person
Heavy goods vehicle
Highly sleep disturbed person
Ilots Regroupés pour l'Information Statistique (Aggregated Units for Statistical Information)
Life cycle assessment
Life cycle inventory
Life cycle impact assessment
Light vehicle
Noise emission and propagation models
Passenger-kilometre
Surrogate model
Tonne-kilometre
Vehicle-kilometre
World Health Organization
*DALY, HAP and HSDP are acronyms. However, they are also used in this work as units for convenience and consistency. Therefore, it has been chosen not to agree these nouns in number as if they were unit symbols.
viii
Publications
Articles
Meyer, R., Benetto, E., Igos, E., Lavandier, C., 2016. Analysis of the different techniques to include noise damage
in life cycle assessment. A case study for car tires. Int. J. Life Cycle Assess. 1–14.
Meyer, R., Lavandier, C., Gauvreau, B., Benetto, E., 2017. Influence of the search radius in a noise prediction
software on population exposure and human health impact assessments. Appl. Acoust. 127, 63–73.
Oral presentations
Meyer, R., Igos, E., Benetto, E., 2016. Analysis of the different technique to include noise damage in life cycle
assessment. A case study for car tires. Presented at the SETAC Europe 26th Annual Meeting, Nantes, France.
Meyer, R., Lavandier, C., Gauvreau, B., Vincent, B., 2016. Intégration du bruit de trafic routier dans l’analyse du
cycle de vie: influence de la distance de propagation entre sources et habitations sur l’évaluation des populations
exposées, in: CFA 2016, Congrès Français D’acoustique. Université du Maine, le Mans, p. pp–2077.
Posters
Meyer, R., Benetto, E., 2015. Towards a new midpoint indicator for including noise impacts from mobility in
LCA. Presented at the 9th International Conference on Society & Materials (SAM9), Luxembourg.
Meyer, R., Benetto, E., Lavandier, C., 2016. Towards a new midpoint indicator for including noise impacts from
mobility in life cycle assessment. Presented at the SETAC Europe 26th Annual Meeting, Nantes, France.
Meyer, R., Benetto, E., Lavandier, C., 2017. A new method to integrate noise impact from road mobility in life
cycle assessment - preliminary results. Presented at the SETAC Europe 27th Annual Meeting, Brussels, Belgium.
MT180
I have participated to MT180, the French adaptation of Three Minute Thesis (3MT®), a competition where PhD
students have to effectively explain their research in three minutes. My presentation occurred on 25 th May 2017
and I won the public price. Video is available at https://www.youtube.com/watch?v=JKYUAKQ_k_g (first three
minutes).
Videos
Even if it is not directly linked to my PhD thesis, I have realised multiple videos on a Youtube channel Le
Réveilleur. The main discussed topic is environment (climate change, pollution, resource depletion…). In two
years, I have realised more than thirty videos and reached 10k subscribers. I consider that scientific dissemination
is a vital part of my job as a researcher in the environmental field since research in the environmental field has,
by nature, political implications. https://www.youtube.com/c/LeR%C3%A9veilleur.
In particular, in the noise exposure assessment, and thus in all the calculation made during this work, the
population is located as its residential address over the whole day. This hypothesis has been made to simplify the
situation and make this whole modelling possible. However, the real-life situation may significantly differ from
this hypothesis. For example, workers spend a large part of the day period of their week-time at work where their
exposure can be totally different. This assumption makes more sense during the night period where most of the
population stays at home. Correcting this assumption with a better modelling of the population’s location should
affect more the annoyance estimates than sleep disturbance ones.
Tenailleau et al. (2015) showed significant differences in terms of population exposure if the weighted sound level
LAeq,24H was calculated on the façade or on a buffer surrounding the address point. The larger the buffer, the higher
the population exposure. Since a given inhabitant is moving within his neighbourhood, this approach could give
a better idea of the noise at which he is exposed. Moreover, this approach could be implemented in this method
without many complications.
Kaddoura et al. (2017) used an activity-based dynamic approach to analyse population exposure to road traffic
noise using the activity-based open-source simulation framework MATSim (http://www.matsim.org/). Daily
travel plans were generated for each agent, and noise emissions and immissions were calculated per hour.
92
Population exposure was calculated under two different assumptions. In the first case, damage costs are only
incurred for residents who are at home, and the noise to which individuals are exposed while performing another
activity type is assumed not to cause any damages. In a second case, noise damage costs are incurred for residents
being at home, at work or at their place of education. The difference between these two assumptions can be high
(> +100% during daytime). However, to understand the implication of assuming population at home during the
whole day, it would be more informative to first perform a noise damage assessment assuming that the whole
population was at their residence address compared to the second case of Kaddoura et al. (2017), where other
activity types are considered. Kaddoura et al. (2017) do not quantify the possible errors in the noise exposure
assessment induced by the assumption of locating the entire population at home during the whole day. Kaddoura
et al. (2017) also demonstrated the power of an activity-based approach to assess population exposure, especially
if one wants to assess the within-day dynamics of noise exposure and population density within the city.
Another possibility, simpler to implement, could be to allocate the population at their workplace or place of
education during a percentage of the day period.
Any attempt to better model the population location could significantly modified the CFs produced with this
method, increasing their representativeness, accuracy and reliability. These limits and warnings are not specific
to this work since all methods relying on population exposure data rely indirectly on the same assumptions.
7.3.4 Incomplete uncertainty quantification
The spatial variability (part of the ontic uncertainties, i.e. the inherent variability of the studied phenomenon) was
analysed in this work, and a quantification has been given for the sample used in this work. For the epistemic
uncertainties (coming from errors and approximations in the models, in the noise prediction software and input
data), only a rough qualitative sensitivity analysis have been done to understand the potential impact of different
error sources. Several points have not been evaluated: the difference between emission and propagation models
existing in the acoustic world and thus the uncertainty coming with the choice of one of this model; or the
uncertainty coming from the implementation in the noise prediction software used in this work. For example,
uncertainty coming with the assumption discussed before (7.3.3) has not been quantified and this quantification
may be difficult. A systematic uncertainty quantification would need further work, repetitions of the method with
other approaches, models and tools. It may be possible that a good understanding of all sources of uncertainties
need the development of an open-source software. The inability to correctly assess all sources of uncertainties is
a limit of this work.
Uncertainty quantification has been provided for the spatial variability of the sample used in this work. It is already
an improvement compared to existing works which integrate environmental noise impacts on human health that
offer no uncertainty quantification, or at best, few values calculated in different situations.
7.4 Using the IRIS division of the territory as basis for the calculation
This work used an existing geographical division: the IRIS used by the French National Institute of Statistics and
Economic Studies (INSEE). This geographical division ensures homogeneity among geographic and demographic
criteria. Thus, this choice of geographical division should have made the identification of the potential typology
easier. However, it was not possible to establish a typology in the studied sample. The only reason to use the IRIS
93
geographical division then fades away. The methodology was thought to avoid mistakes while assessing the
marginal CFs, so these should not have been impacted by the varying size of the IRIS. In contrast, average CFs
are sensitive to the surroundings of the area in which they are calculated, and using geographical subunits of
different size and shape may affect average CFs. From a methodological point of view, practitioners could use a
normalized grid to calculate CFs. Doing so would randomize the division in geographical subunits to analyse and
avoid unpredictable effects coming from existing geographical subunits.
7.5 Pertinence of spatialized CFs for the environmental noise question
Noise is a peculiar impact category for LCA because of its physical nature and its localisation in both time and
space. Noise fades away in seconds and propagates at several kilometres in the extreme cases. To overcome some
of these problems, environmental noise assessment considers noise emission and propagation as averages over
long periods. CFs calculated in two different areas, separated by several hundred of meters, can change by an
order of magnitude. Since the object under study is a vehicle, it is safe to assume that the emission point of the
sound energy will travel in areas with different CF values. Considering both the specific physical nature of sound
and the moving point of the emission, it makes more sense to use weighted mean CFs than to spatially differentiate
it. This is the proposed approach in this work since we were not able to produce a clear, usable and statistically
justifiable typology.
Despite not finding one here, a typology can still exist. First, the studied sample is almost entirely constituted by
IRIS with a density greater than 1000 inhabitants per km². If a typology exists for low density areas or between
high density areas and the low density areas, which are not considered in this work, it was missed because of the
sampling. It is difficult to guess what would be the CFs for low density areas. If the SMs found in this work still
apply in low density areas, a decrease in population density can be compensated by a decrease in sound energy
density. Additionally, average CFs are higher than marginal CFs. If marginal and average CFs are correctly
evaluated, CFs in low density areas can thus be higher than the ones found in this work for higher density. It is
important to note that this is purely speculative, and additional work should focus on low density areas to have a
clear view of what happens to the CFs in such an environment.
A typology may also exist based on the driving behaviours. Energy-based CFs for LVs and HGVs present
differences because of differences of the repartition of these two vehicle types on the road network. Since the
amount of traffic existing on a road is a key parameter for the marginal CFs. It may be interesting to evaluate CFs
for different road types and then establish behavioural profile for different types of transportation (at least for
freight and passenger). Since Franco et al. (2010) and Moliner et al. (2014) found low values while looking at
motorways, this may likely be a specific case for the environmental noise impact of road transportation (since low
population density and high traffic leads both to lower CFs).
Finally, typology can still exist on a wider scale. Doing a series of calculations on normalized grid of different
sizes may reveal a typology at one scale while it was impossible to detect one at smaller or higher scale. A practical
way to do this experiment would be (if possible in terms of time and calculation) to calculate all the variables on
a small grid (500m x 500m for example) and then to calculate CFs by aggregating more and more subunits. CFs
would be calculated on 500m x 500m, then 1km x 1km, then 2km x 2km, etc… A typology may appear for a
94
given size of the calculation area if the averaging on a higher surface masks the spatial variability over short
distances found in this work.
7.6 Distance-based approach and vehicle categories
The multiple CFs calculated in this work helped us to better understand the environmental noise impacts on human
health and the best way to integrate it in LCA. Marginal CFs were calculated with both distance-based and energy-
based approaches. A distance-based approach was the natural way to follow previous work aiming to integrate
noise impacts in LCA (Franco et al., 2010; Moliner et al., 2014; Müller-Wenk, 2004) and existing LCIA
methodologies (Frischknecht and Büsser Knöpfel, 2013; Steen, 1999a). Calculating these CFs allowed the
comparison of values found in this work with existing ones. An energy-based approach fits better with the general
LCA philosophy by relying on an EF in terms of energy, and this was already proposed by Cucurachi et al.(2012).
Examining both marginal and average CFs allows to compare these approaches and to have a clearer view of
advantages and drawbacks coming with the choice of the EF. At the term of this work, energy-based CFs seem
more logical and more reliable to quantify more accurately the sound energy and thus the physical stressor behind
environmental noise impacts on human health. Energy-based CFs fit better in the LCA framework and allow the
aggregation of traffic from different vehicle types. Since noise emission power must be calculated for each process
in order to quantify the outgoing EF in joule, it may be more challenging to develop the LCI processes. However,
a more refined distinction of noise emission in the LCI would lead to a more accurate LCIA.
The CNOSSOS-EU noise assessment method (Kephalopoulos et al., 2012) is the common method to be used by
EU Members to apply the Environmental Noise Directive 2002/49/EC (Directive, 2002). For the propagation part,
this method is mostly based on, and is similar to, the NMPB 2008 method (SETRA Copyright (Collective), 2009b)
used in this work. However, the emission part takes into consideration more vehicle categories: it divides the
traffic in at least four categories: light motor vehicles, medium heavy vehicles, heavy vehicles and powered two-
wheelers. This subdivision allows a more accurate calculation of the noise emission. In the available data, there
were only two vehicle categories, as described in SETRA Copyright (Collective) (2009a). Noise mapping on the
European scale should apply the CNOSSOS-EU model. Cities calculating noise maps should then collect data
corresponding to the four vehicle categories listed above if they are not already doing so.
Distance-based CFs should be elaborated for each new vehicle type since there is a significant difference in terms
of emission. On the other hand, if two different vehicle types are distributed in a similar way over the road network,
and if they are considered equivalent sources for assessing damage to human health (i.e. have the same dose-
response relationships), the energy-based CFs should be identical. Energy-based CFs can then be a simple solution
to avoid complications while refining the emission model by adding new vehicle types. Since the emission part is
integrated in the LCI process (when calculating the noise emission power to establish the output EF in joule),
energy-based CFs are not dependent on the difference in terms of emissions of the different vehicle types. Energy-
based CFs can thus considerably simplify the integration of noise impacts on human health for all the
transportation processes in LCA.
Moliner et al. (2013) proposed a method to calculate the external costs of road traffic noise in which the noise
cost allocation between different vehicle categories is based on the noise emission power. This approach is similar
95
to the one proposed to refine the inventory by differentiating the inventory based on the noise emission power of
the different vehicles.
7.7 Use of marginal and average CFs
Average CFs were calculated here, not only because it was possible and advised by UNEP (2016) but also because
it is thought to be informative. In particular, comparing average CFs to marginal ones provided several pieces of
information. Marginal and average CFs close in value indicate a low sensitivity to the traffic in the reference
situation in which they are calculated and thus a possibility to use them for a large range of traffic variations. But,
under which conditions could these average CFs related to noise be used in an LCA? As a first approximation,
differentiating between the use of marginal CFs or average CFs is dependent on the question one wants to answer:
If one wants to know what will be the consequences of a policy, a choice or a change, marginal CFs are
advised.
If one wants to know what is the impact of an existing product, service or process in the ongoing economy,
average CFs should be used.
A special case is the assessment of a significant change where marginal CFs may not be valid anymore. In this
case, it would be better to specifically calculate the impact resulting from this change without using CFs, especially
if the impact category under consideration presents thresholds and/or non-linear behaviour. Having marginal and
average CFs significantly different for a given impact category can be a good indication that said impact category
needs a specific modelling when a large change is occurring.
For noise impacts of road transportation on human health, in case of a significant decrease in the road
transportation noise, using average CFs instead of marginal ones can be preferred. For example, to evaluate the
following policy “reducing road traffic by 90% in a given area”, even if a change is under study, using average
CFs would provide more reliable results because this change is far from being marginal. This specific case is thus
an exception of the rules given above (assessing a consequence with marginal CFs) because the average CFs are,
in this case, a better proxy for the change. However, the best solution for assessing non-marginal change will
always be a specific modelling of the situation since the calculated CFs do not apply.
Consideration between attributional and consequential LCI does not superpose the consideration of marginal or
average CFs, and various combinations can be thought of:
An assessment of a significant reduction change would preferentially use a consequential LCI, because said
change is most likely affecting other economic activities and average CFs, as explained before. This case is
the most disputable one because the average CFs are considered to be more representative than marginal CFs
for a large change. A specific modelling and personalised CFs would be preferred.
An assessment of a marginal change could use a consequential LCI and marginal CFs. For example, adding
additional electric vehicles on the road network would imply changes in the economy, such as an increased
demand for electricity or new infrastructure to recharge the electric cars (consequential LCI). In this case,
additional noise emissions are added to the existing situation (marginal CFs for environmental noise).
96
An environmental product declaration (EPD) could use attributional LCA to simplify the collection of the
inventory for a product that has minor effects on the rest of the economy or is already being produced. This
EPD should use marginal CFs because it aims to inform the consumer of the impact of his choice and thus of
the consequence of his action.
Calculation of a footprint should use an attributional LCI and average CFs, such as if one wants to develop a
noise footprint for a given population. The rationale should be to allocate the overall noise emissions
(preferentially quantified on an energy basis) to the concerned population, and then to multiply the found
amounts by average CFs. Using marginal CFs to establish a footprint could lead to uninterpretable results,
like the sum of all the footprints not being equal to the overall impact. Using marginal CFs instead of average
ones to establish footprints would lead to an overestimation if the marginal CFs are higher than the average
ones, or an underestimation otherwise.
The example of the footprint and the importance of having a sum corresponding to the overall impact justifies the
average approach considering the whole impact in opposition to the “zero-effect” approach (section 5.2.2 and
Figure 2). Using the “zero-effect” approach would lead to errors.
Marginal CFs are preferred in most cases in LCA, since most LCAs study a marginal change in an ongoing
economy. However, in some specific cases, it is advised to use average CFs. This particularly holds for a
significant change (since, in this case, average CFs may be more representative of the change than the marginal
ones) and for accounting (noise footprint, reporting on prior policies).
In addition, the average CFs developed in this work must be used cautiously; they are more uncertain than the
marginal ones for two reasons. First, contrary to marginal CFs, they are subject to some biases in the calculation
(influence of the surroundings) that are not avoided by the method developed in this work. This method focuses
on minimizing errors while calculating marginal CFs. Second, average CFs are more sensitive to potential errors
related to the population exposure assessment and are thus more uncertain.
7.8 Temporal differentiation of environmental noise impact on human health
There are significant differences between CFs for day and for night. However, it may be difficult to temporally
differentiate the LCI at such a small scale. A temporal differentiation of the LCI should be possible by proposing
different EFs for the night and day periods for each road transportation process as proposed in this work.
Temporal differentiation of the impacts would show weekly and monthly pattern since the traffic between
workday and week-end, and it could be really useful to determine traffic between seasons, especially in seasonal
touristic areas. This aspect is out of the scope of this work since noise exposure assessment is based on an average
traffic and thus exposure over a year. Each day is considered identical, so weekly and monthly pattern are thus
masked. Temporal variation among a given period (day or night) also exists. This is evident when considering the
congestion in major cities that coincides with rush hours. Studying these questions may be difficult if the
knowledge and tools from the acoustic world are not yet available. These refinements in the noise exposure
assessment seem out of the reach of LCA with the exception of a specific modelling of traffic in the foreground.
Temporally differentiating the LCI by proposing new EFs for every hour, every week or month, etc. would be
impractical. However, if the occurrence of a given transport is accurately defined in time with an approach like
97
the one developed by Tiruta-Barna et al. (2016), it may be possible to define a temporal function that adjusts the
CFs to integrate monthly, weekly or hourly patterns.
7.9 Uncertainty of environmental noise impact
The quantification of environmental noise impacts is uncertain. Means are used throughout the quantification
process: when calculating the sound energy EF in the LCI; applying CFs; considering environmental noise
exposure as an average level over given periods; etc… The acoustic and epidemiologic knowledge on which this
work is built is also full of uncertainties, from the differences between real-life noise characteristics (emission and
propagation) and the ones modelled to the large uncertainties related to DWs.
Using a holistic approach such as LCA for an instantaneous and short-distance phenomenon such as noise will
inevitably include a certain level of uncertainty. The problem of this uncertainty occurs when aggregating together
damage coming from impact categories with different uncertainties at the endpoint level. Different impact
categories (e.g. climate change or ozone depletion) have different cause-effect chains and thus different
uncertainties. It would be better to quantify uncertainties of the different impact categories prior to aggregation to
better assess the reliability of the results. Saying that the impacts of environmental noise on human health is higher
than the ones coming from climate change for a given product or process may be a disputable statement without
uncertainty quantification of these two different impact categories.
The choice of modelling spatial variability with an uncertainty distribution is scientifically sound, but the effort
may be futile if other CF developers do not do the same. The result of an LCA should always provide uncertainty
quantification since LCA results are nearly impossible to verify in practice. However, not all CFs for the various
impact categories are given with uncertainty quantification, and LCA software does not always allow the
conduction of an uncertainty analysis that includes the uncertainty distribution for the CFs. The integration of
impacts tainted with high uncertainties into LCIA, such as environmental noise impact on human health, relies on
a more accurate assessment of the uncertainties in the LCA framework.
98
8 Conclusion and outlook
There are sufficient elements in the acoustic and epidemiologic literature to calculate the impacts of noise from
road transportation on human health. Not only do these elements exist, but they are also available in a suitable
format to allow for their integration into LCA. Several attempts to solve this question already existed in the LCA
literature prior to this thesis. Applying these existing methods on a case study allowed the understanding of the
pros and cons as well as the reliability of existing approaches. This thesis adopts a third approach that conceptually
fits between the two main existing ones. This third approach consists of using noise emission and propagation
models to accurately assess the road transportation noise impacts on human health and, in turn, to generate CFs
that could be systematically and straightforwardly used in LCIA.
An EF in joules is more natural for the LCA framework, fitting well the definition of EF and the general approach
used in the LCI. Since the noise emission power level per vkm is not constant and varies with different parameters,
such as the speed, a distance-based approach poorly approximates the physical stressor causing human health
damage, i.e. sound energy. Energy-based CFs facilitate a finer characterisation at the LCI stage, leading to a more
accurate and reliable quantification of environmental noise impacts on human health. This work advocates for the
use of sound energy as an EF.
Significant differences were found between CFs for the day period, related to annoyance, and CFs for the night
period, mostly related to sleep disturbance. Temporal variation on increments shorter than a day may be difficult
to assess in LCA, a discipline wherein a fine temporal differentiation is not common. The difference between CFs
for the day and night periods may justify the additional effort to produce a temporal differentiation in the LCI.
Establishing a typology and elaborating CFs for this typology was perceived as the best way to take into account
the spatial variability that potentially exists between different geographical situations. However, it was not
possible to identify a typology in this work. It was found that CFs vary over a small distance. The source under
consideration, a vehicle, is a moving source, and the average travel distance is substantial in comparison to the
distance under which CFs vary. A statistical approach approximating the noise impact of a given vehicle with an
uncertainty distribution taking into account the spatial variability is thus a more logical, reliable and practical
approach. These uncertainty distributions could be used by practitioners to propagate uncertainty and take into
account the inherent ontic uncertainties of the environmental noise impact on human health. A practical use of
these uncertainty distributions were proposed, and it demonstrated the feasibility and the utility of this approach.
These uncertainty distributions allow to take into account the inherent variability of this impact due to the physical
characteristics of noise. Noise will always be localised in both time and space at a scale smaller than the majority
of other impacts integrated in LCA. The inability to reduce the ontic uncertainty of noise impacts advocates for a
better integration of uncertainty in the whole LCA framework, especially in LCIA where it is still far from
common practice.
Average CFs can be useful in specific and limited situations, such as a significant reduction in road traffic or the
elaboration of a noise footprint. Producing average CFs in this work allowed for a comparison with marginal CFs,
which showed a low sensitivity of the CFs towards the traffic in the reference situation in which they were
calculated. Producing average and marginal CFs was shown to be useful from the modeller perspective.
99
This work proves the utility of existing noise emission and propagation models as powerful tools to produce CFs.
Noise prediction software, acoustic models and availability of the data will continue to improve. The development
of emission and propagation models, their implementation in noise prediction software and the studies of
environmental noise impact on human health are still open and very active research questions. Developments in
the upcoming years will improve the tools and knowledge underlying both the method developed here and the
integration of environmental noise in LCA in general. Open-source noise prediction software could offer good
opportunities for systematic and streamlined calculations of CFs in a near future. This would allow a better
mastery of the different calculation aspects, thereby improving uncertainty estimation.
It is imbalanced to produce CFs for road transportation processes while not doing it for other transportation
processes. This work focuses on road transportation processes, but the method and approaches produced here
could be reproduced for other transportation means, especially for trains and planes where both data and dose-
response relationships already exist. Other transportation processes may bring new challenges, but the results
would yield a fairer LCA outcome for the different transportation modes.
Integrating the CFs found in this work in existing road transportation processes could show a doubling of the
damage on human health. This high human health damage is aligned with the existing literature, even if this
magnitude of damage can be surprising. This underlines the necessity to integrate environmental noise impacts
on human health in LCIA. A holistic approach cannot neglect an impact of this importance without raising
criticism and without leading to potentially large errors. This result should also bring interest and further research
effort in the modelling of this impact category.
To a large extent, results of this work are dependent on the initial choice to involve, via the steering committee,
experts both from the LCA and acoustic communities. Integration of a new impact in LCIA may only be done
properly with the help of experts of the field under consideration. It is not always easy for two different worlds
with different thinking, approaches and lexicons to talk to each other, but this work proves the dialogue is possible.
It was rewarding, and it produced the foundation of this work. Here is to hoping that similar multidisciplinary
approaches will continue to be investigated, either for further work on environmental noise, or for the integrations
of other impacts in general.
100
9 References Abatement, U.S.O. of N., Control, 1974. Information on levels of environmental noise requisite to protect public
health and welfare with an adequate margin of safety. for sale by the Supt. of Docs., US Govt. Print. Off. AFNOR, 2011. Norme NF S31-133, Acoustique des milieux extérieurs - Calcul des niveaux sonores. ALAYRAC, M., 2009. Indicateurs de gêne sonore pour l’étude d’impact du bruit d’un site industriel:
caractérisation physique et perceptive. Ecole Nationale des Travaux Publics de l’Etat. Althaus, H.-J., Haan, P. de, Scholz, R.W., 2009. Traffic noise in LCA. Int. J. Life Cycle Assess. 14, 560–570.
https://doi.org/10.1007/s11367-009-0116-2 Andy Field, Z., Miles, J., 2012. Discovering statistics using R. Sage, Thousand Oaks. Aumond, P., Can, A., De Coensel, B., Botteldooren, D., Ribeiro, C., Lavandier, C., 2017. Modeling Soundscape
Pleasantness Using perceptual Assessments and Acoustic Measurements Along Paths in Urban Context. Acta Acust. United Acust. 103, 430–443.
Babisch, W., others, 2008. Road traffic noise and cardiovascular risk. Noise Health 10, 27. Babisch, W., Pershagen, G., Selander, J., Houthuijs, D., Breugelmans, O., Cadum, E., Vigna-Taglianti, F.,
Katsouyanni, K., Haralabidis, A.S., Dimakopoulou, K., Sourtzi, P., Floud, S., Hansell, A.L., 2013. Noise annoyance — A modifier of the association between noise level and cardiovascular health? Sci. Total Environ. 452, 50–57. https://doi.org/10.1016/j.scitotenv.2013.02.034
Babisch, W., Van Kamp, I., others, 2009. Exposure-response relationship of the association between aircraft noise and the risk of hypertension. Noise Health 11, 161.
Bare, J.C., Hofstetter, P., Pennington, D.W., Haes, H.A.U. de, 2000. Midpoints versus endpoints: The sacrifices and benefits. Int. J. Life Cycle Assess. 5, 319. https://doi.org/10.1007/BF02978665
Berglund, B., Berglund, U., Goldstein, M., Lindvall, T., 1981. Loudness (or annoyance) summation of combined community noises. J. Acoust. Soc. Am. 70, 1628–1634.
Berglund, B., Lindvall, T., Schwela, D.H., others, 1999. Guidelines for community noise, in: Guidelines for Community Noise. OMS.
Botteldooren, D., Verkeyn, A., 2002. Fuzzy models for accumulation of reported community noise annoyance from combined sources. J. Acoust. Soc. Am. 112, 1496–1508. https://doi.org/10.1121/1.1506366
Carnell, R., 2016. triangle: Provides the Standard Distribution Functions for the Triangle Distribution. Carson, R., 1962. Silent spring. Houghton Mifflin Harcourt. Clark, C., Head, J., Stansfeld, S.A., 2013. Longitudinal effects of aircraft noise exposure on children’s health and
cognition: A six-year follow-up of the UK RANCH cohort. J. Environ. Psychol. 35, 1–9. https://doi.org/10.1016/j.jenvp.2013.03.002
Cucurachi, S., Heijungs, R., 2014. Characterisation factors for life cycle impact assessment of sound emissions. Sci. Total Environ. 468–469, 280–291. https://doi.org/10.1016/j.scitotenv.2013.07.080
Cucurachi, S., Heijungs, R., Ohlau, K., 2012. Towards a general framework for including noise impacts in LCA. Int. J. Life Cycle Assess. 17, 471–487. https://doi.org/10.1007/s11367-011-0377-4
Cucurachi, S., Heijungs, R., Peijnenburg, W.J.G.M., Bolte, J.F.B., de Snoo, G.R., 2014. A framework for deciding on the inclusion of emerging impacts in life cycle impact assessment. J. Clean. Prod. 78, 152–163. https://doi.org/10.1016/j.jclepro.2014.05.010
Curran, M.A., Mann, M., Norris, G., 2005. The international workshop on electricity data for life cycle inventories. J. Clean. Prod. 13, 853–862.
De Coensel, B., Sun, K., Wei, W., Van Renterghem, T., Sineau, M., Ribeiro, C., Can, A., Aumond, P., Lavandier, C., Botteldooren, D., 2015. Dynamic noise mapping based on fixed and mobile sound measurements, in: EURONOISE 2015. p. 6p.
Delignette-Muller, M.L., Dutang, C., 2015. fitdistrplus: An R Package for Fitting Distributions. J. Stat. Softw. 64, 1–34.
Directive, E., 2002. Directive 2002/49/EC of the European parliament and the Council of 25 June 2002 relating to the assessment and management of environmental noise. Off. J. Eur. Communities 189.
Doka, G., 2003. Ergänzung der Gewichtungsmethode für Ökobilanzen Umweltbelastungspunkte’ 97 zu Mobilitäts-UBP’ 97. Doka Ökobilanzen Zurich.
Drott, E., Bruce, R., 2011. A different look at noise exposure, hearing loss, and time limits. Hear Rev 18, 34–36. Dutilleux, G., Defrance, J., Ecotière, D., Gauvreau, B., Bérengier, M., Besnard, F., Duc, E.L., 2010. NMPB-
Routes-2008: The Revision of the French Method for Road Traffic Noise Prediction. Acta Acust. United Acust. 96, 452–462. https://doi.org/10.3813/AAA.918298
Ecotiere, D., Foy, C., Dutilleux, G., 2012. Comparison of engineering models of outdoor sound propagation: NMPB2008 and Harmonoise-Imagine, in: Acoustics 2012.
Ekvall, T., Azapagic, A., Finnveden, G., Rydberg, T., Weidema, B.P., Zamagni, A., 2016. Attributional and consequential LCA in the ILCD handbook. Int. J. Life Cycle Assess. 21, 293–296. https://doi.org/10.1007/s11367-015-1026-0
101
European Commission, 2011. European Commission - Joint Research Centre - Institute for Environment and Sustainability: International Reference Life Cycle Data System (ILCD) Handbook – Recommendations for Life Cycle Impact Assessment in the European context – based on existing environmental impact assessment models and factors., First edition. ed. Luxembourg. Publications Office of the European Union.
Fidell, S., Mestre, V., Schomer, P., Berry, B., Gjestland, T., Vallet, M., Reid, T., 2011. A first-principles model for estimating the prevalence of annoyance with aircraft noise exposure. J. Acoust. Soc. Am. 130, 791–806.
Finnveden, G., Hauschild, M.Z., Ekvall, T., Guinée, J., Heijungs, R., Hellweg, S., Koehler, A., Pennington, D., Suh, S., 2009. Recent developments in life cycle assessment. J. Environ. Manage. 91, 1–21.
Flash Eurobaromètre, 2010. Enquête d’opinion sur la qualité de la vie dans 75 villes européennes. Franco, V., Garraín, D., Vidal, R., 2010. Methodological proposals for improved assessments of the impact of
traffic noise upon human health. Int. J. Life Cycle Assess. 15, 869–882. https://doi.org/10.1007/s11367-010-0213-2
Frischknecht, R., Büsser Knöpfel, S., 2013. Swiss eco-factors 2013 according to the ecological scarcity method. Methodol. Fundam. Their Appl. Switz. Environ. Stud. 254.
Frischknecht, R., Jungbluth, N., Althaus, H.-J., Doka, G., Dones, R., Heck, T., Hellweg, S., Hischier, R., Nemecek, T., Rebitzer, G., others, 2005. The ecoinvent database: Overview and methodological framework (7 pp). Int. J. Life Cycle Assess. 10, 3–9.
Fritschi, L., Brown, L., Kim, R., Schwela, D., Kephalopolous, S., 2011. Conclusions [Burden of disease from environmental noise: Quantification of healthy years life lost in Europe]. World Health Organisation.
Fyhri, A., Klæboe, R., 2009. Road traffic noise, sensitivity, annoyance and self-reported health—A structural equation model exercise. Environ. Int. 35, 91–97.
GaBi, 2015. Life Cycle Engineering software and database. LBP, University of Stuttgart and thinkstep. Galbrun, L., Ali, T.T., 2013. Acoustical and perceptual assessment of water sounds and their use over road traffic
noise. J. Acoust. Soc. Am. 133, 227–237. https://doi.org/10.1121/1.4770242 Goedkoop, M., Heijungs, R., Huijbregts, M., De Schryver, A., Struijs, J., van Zelm, R., 2009. ReCiPe 2008. Life
mapping based on participative measurements. Noise Mapp 3, 140–156. Guski, R., others, 1999. Personal and social variables as co-determinants of noise annoyance. Noise Health 1, 45. Haagsma, J.A., Polinder, S., Cassini, A., Colzani, E., Havelaar, A.H., 2014. Review of disability weight studies:
comparison of methodological choices and values. Popul. Health Metr. 12, 20. https://doi.org/10.1186/s12963-014-0020-2
Hansell, A.L., Blangiardo, M., Fortunato, L., Floud, S., de Hoogh, K., Fecht, D., Ghosh, R.E., Laszlo, H.E., Pearson, C., Beale, L., others, 2013. Aircraft noise and cardiovascular disease near Heathrow airport in London: small area study. Bmj 347, f5432.
Health Council of the Netherlands, 1971. Committee on Noise Annoyance and noise abatement. Geluidhinder [Noise annoyance, in Dutch].
Hellweg, S., Milà i Canals, L., 2014. Emerging approaches, challenges and opportunities in life cycle assessment. Science 344, 1109–1113.
Heutschi, K., des forêts et du paysage Suisse. Office fédéral de l’environnement, Eidgenössische Materialprüfungs-und Forschungsanstalt. Akustik, Geinoz, D., Liengme, J.-D., 2004. SonRoad: modèle de calcul du trafic routier. OFEFP.
http://www.matsim.org/, n.d. MATSim - Multi-Agent Transport Simulation. Huijbregts, M.A., Hellweg, S., Hertwich, E., 2011. Do we need a paradigm shift in life cycle impact assessment?
Environ. Sci. Technol.-Columb. 45, 3833. Huijbregts, M.A.J., Steinmann, Z.J.N., Elshout, P.M.F., Stam, G., Verones, F., Vieira, M., Zijp, M., Hollander,
A., Zelm, R. van, 2017. ReCiPe2016: a harmonised life cycle impact assessment method at midpoint and endpoint level. Int. J. Life Cycle Assess. 22, 138–147. https://doi.org/10.1007/s11367-016-1246-y
Ifop, 2014. Les français et les nuisances sonores [WWW Document]. URL http://www.bruit.fr/images/stories/pdf/sondage-ifop-2014-synthese.pdf (accessed 8.8.17).
IMAGINE, 2007a. Guidelines for the use of traffic models for noise mapping and noise action planning. Deliverable 7 of the IMAGINE (Improved methods for the assessment of the generic impact of noise in the environment) Project, TNO Technical Report, Delft, The Netherlands.
IMAGINE, 2007b. The noise emission model for European road traffic. Deliverable 11 of the IMAGINE (Improved methods for the assessment of the generic impact of noise in the environment) Project, M+P—Consulting Engineers Technical Report, Vught, The Netherlands.
Institut national de la statistique et des études économiques (INSEE) [WWW Document], n.d. URL http://www.insee.fr/fr/ (accessed 2.26.16).
International Standard IEC 61672-1, 2003. Electroacoustics – Sound level meters – Part 1: Specifications. ISO 3744, 2010. Acoustics — Determination of sound power levels and sound energy levels of noise sources
using sound pressure — Engineering methods for an essentially free field over a reflecting plane [WWW Document]. URL https://www.iso.org/obp/ui/#iso:std:iso:3744:ed-3:v1:en (accessed 10.11.16).
ISO 14040, 2006. Environmental management -- Life cycle assessment -- Principles and framework. Environ. Manag.-Life Cycle Assess.-Requir. Guidel. Eur. Comm. Stand.
ISO 14044, 2006. Environmental management -- Life cycle assessment -- Requirements and guidelines. Environ. Manag.-Life Cycle Assess.-Requir. Guidel. Eur. Comm. Stand.
Jolliet, O., Margni, M., Charles, R., Humbert, S., Payet, J., Rebitzer, G., Rosenbaum, R., 2003. IMPACT 2002+: a new life cycle impact assessment methodology. Int. J. Life Cycle Assess. 8, 324–330.
JRC, 2016. European Life Cycle Database. European Comission. Kaddoura, I., Kröger, L., Nagel, K., 2017. An activity-based and dynamic approach to calculate road traffic noise
damages. Transp. Res. Part Transp. Environ. 54, 335–347. Kephalopoulos, S., Paviotti, M., Ledee, F.A., 2012. Common noise assessment methods in Europe (CNOSSOS-
EU). Publications Office of the European Union. King, E.A., Murphy, E., Rice, H.J., 2011. Implementation of the EU environmental noise directive: Lessons from
the first phase of strategic noise mapping and action planning in Ireland. J. Environ. Manage. 92, 756–764.
Kroesen, M., Molin, E.J., van Wee, B., 2008. Testing a theory of aircraft noise annoyance: A structural equation analysis. J. Acoust. Soc. Am. 123, 4250–4260.
Levain, J.-P., Mauny, F., Pujol, S., Petit, R., Houot, H., Defrance, J., Lardies, J., Berthillier, M., 2015. Exposition au bruit et performance scolaire des élèves de CE2. Psychol. Fr. 60, 35–49. https://doi.org/10.1016/j.psfr.2014.07.001
Lopez, A.D., others, 2006. Global burden of disease and risk factors. World Bank Publications. Luxon, L.M., Prasher, D., 2007. Noise and its effects. John Wiley & Sons. Marquis-Favre, C., Premat, E., Aubrée, D., 2005. Noise and its effects–a review on qualitative aspects of sound.
Part II: Noise and annoyance. Acta Acust. United Acust. 91, 626–642. Meyer, R., Benetto, E., Igos, E., Lavandier, C., 2016. Analysis of the different techniques to include noise damage
in life cycle assessment. A case study for car tires. Int. J. Life Cycle Assess. 1–14. Meyer, R., Lavandier, C., Gauvreau, B., Benetto, E., 2017. Influence of the search radius in a noise prediction
software on population exposure and human health impact assessments. Appl. Acoust. 127, 63–73. Miedema, H.M.E., Vos, H., 2007. Associations between self-reported sleep disturbance and environmental noise
based on reanalyses of pooled data from 24 studies. Behav. Sleep. Med. 5, 1–20. Miedema, H.M., Oudshoorn, C.G., 2001. Annoyance from transportation noise: relationships with exposure
metrics DNL and DENL and their confidence intervals. Environ. Health Perspect. 109, 409. Miedema, H.M., Passchier-Vermeer, W., Vos, H., 2002. Elements for a position paper on night-time
transportation noise and sleep disturbance. TNO Inro. Mioduszewski, P., Ejsmont, J.A., Grabowski, J., Karpiński, D., 2011. Noise map validation by continuous noise
monitoring. Appl. Acoust., Noise Mapping 72, 582–589. https://doi.org/10.1016/j.apacoust.2011.01.012 Moliner, E., Vidal, R., Franco, V., 2013. A fair method for the calculation of the external costs of road traffic
noise according to the Eurovignette Directive. Transp. Res. Part Transp. Environ. 24, 52–61. https://doi.org/10.1016/j.trd.2013.05.007
Moliner, E., Vidal, R., Franco, V., Garraín, D., 2014. Método para evaluar el impacto del ruido del transporte rodado en el marco del análisis del ciclo de vida. DYNA-Ing. E Ind. 89.
Müller-Wenk, R., 2004. A method to include in lca road traffic noise and its health effects. Int. J. Life Cycle Assess. 9, 76–85. https://doi.org/10.1007/BF02978566
Müller-Wenk, R., 2002. Attribution to road traffic of the impact of noise on health environmental series No. 339: Noise. Int. J. Life Cycle Assess. 7, 376–376. https://doi.org/10.1007/BF02978688
Müller-Wenk, R., 1999. Life-cycle impact assessment of road transport noise. Institut für Wirtschaft und Ökologie, Universität St. Gallen (IWÖ-HSG).
103
Murphy, E., King, E.A., 2010. Strategic environmental noise mapping: Methodological issues concerning the implementation of the EU Environmental Noise Directive and their policy implications. Environ. Int. 36, 290–298.
Murray, C.J., 1994. Quantifying the burden of disease: the technical basis for disability-adjusted life years. Bull. World Health Organ. 72, 429–445.
Nielsen, P., Laursen, J., 2005. Integration of external noise nuisance from road and rail transportation in lifecycle assessment. Dan. Environ. Prot. Agency Cph.
Ongel, A., 2016. Inclusion of Noise in Environmental Assessment of Road Transportation. Environ. Model. Assess. 21, 181–192.
Ongel, A., Sezgin, F., 2016. Assessing the effects of noise abatement measures on health risks: A case study in Istanbul. Environ. Impact Assess. Rev. 56, 180–187.
Pennington, D.W., Potting, J., Finnveden, G., Lindeijer, E., Jolliet, O., Rydberg, T., Rebitzer, G., 2004. Life cycle assessment Part 2: Current impact assessment practice. Environ. Int. 30, 721–739. https://doi.org/10.1016/j.envint.2003.12.009
Picaut, J., Aumond, P., Can, A., Fortin, N., Gauvreau, B., Bocher, E., Palominos, S., Petit, G., Guillaume, G., 2017. Noise mapping based on participative measurements with a smartphone. J. Acoust. Soc. Am. 141, 3808–3808.
R Core Team, 2015. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.
Reported data on noise exposure covered by Directive 2002/49/EC [WWW Document], n.d. . Eur. Environ. Agency. URL https://www.eea.europa.eu/data-and-maps/data/data-on-noise-exposure-2 (accessed 7.6.17).
Ronnebaum, T., Schulte-Fortkamp, B., Weber, R., 1996. Synergetic effects of noise from different sources: a literature study, in: International Congress on Noise Control Engineering. pp. 2241–2246.
Solberg, S., 1996. Total noise rating at Oslo airport Gardermoen, in: International Congress on Noise Control Engineering. pp. 2501–2504.
Sonnemann, G., Vigon, B., others, 2011. Global guidance principles for Life Cycle Assessment (LCA) databases: a basis for greener processes and products.
Sørensen, M., Andersen, Z.J., Nordsborg, R.B., Jensen, S.S., Lillelund, K.G., Beelen, R., Schmidt, E.B., Tjønneland, A., Overvad, K., Raaschou-Nielsen, O., 2012. Road Traffic Noise and Incident Myocardial Infarction: A Prospective Cohort Study. PLOS ONE 7, e39283. https://doi.org/10.1371/journal.pone.0039283
Stansfeld, S., Clark, C., 2015. Health Effects of Noise Exposure in Children. https://doi.org/10.1007/s40572-015-0044-1
Steen, B., 1999a. A systematic approach to environmental priority strategies in product development (EPS): version 2000-general system characteristics. Centre for Environmental Assessment of Products and Material Systems Gothenburg.
Steen, B., 1999b. A systematic approach to environmental priority strategies in product development (EPS): version 2000-Models and data of the default method. Chalmers tekniska högsk.
Steven, H., 2005. Ermittlung der Geräuschemission von Kfz im Straßenverkehr. TÜV Nord Würselen. Tenailleau, Q.M., Bernard, N., Pujol, S., Houot, H., Joly, D., Mauny, F., 2015. Assessing residential exposure to
urban noise using environmental models: does the size of the local living neighborhood matter? J. Expo. Sci. Environ. Epidemiol. 25, 89.
Tiruta-Barna, L., Pigné, Y., Navarrete Gutiérrez, T., Benetto, E., 2016. Framework and computational tool for the consideration of time dependency in Life Cycle Inventory: proof of concept. J. Clean. Prod. 116, 198–206. https://doi.org/10.1016/j.jclepro.2015.12.049
UNEP, 2016. Global Guidance for Life Cycle Impact Assessment Indicators. United Nations Environment Programme.
Van Kempen, E.E., Kruize, H., Boshuizen, H.C., Ameling, C.B., Staatsen, B.A., de Hollander, A.E., 2002. The association between noise exposure and blood pressure and ischemic heart disease: a meta-analysis. Environ. Health Perspect. 110, 307.
Van Maercke, D., Defrance, J., 2007. Development of an Analytical Model for Outdoor Sound Propagation Within the Harmonoise Project. Acta Acust. United Acust. 93, 201–212.
104
Van Zelm, R., Huijbregts, M.A.J., den Hollander, H.A., van Jaarsveld, H.A., Sauter, F.J., Struijs, J., van Wijnen, H.J., van de Meent, D., 2008. European characterization factors for human health damage of PM10 and ozone in life cycle impact assessment. Atmos. Environ. 42, 441–453. https://doi.org/10.1016/j.atmosenv.2007.09.072
Walker, W.E., Harremoës, P., Rotmans, J., van der Sluijs, J.P., van Asselt, M.B., Janssen, P., Krayer von Krauss, M.P., 2003. Defining uncertainty: a conceptual basis for uncertainty management in model-based decision support. Integr. Assess. 4, 5–17.
Warnes, G.R., Bolker, B., Lumley, T., 2015. gtools: Various R Programming Tools. Watts, G.R., Pheasant, R.J., Horoshenkov, K.V., Ragonesi, L., 2009. Measurement and Subjective Assessment of
Water Generated Sounds. Acta Acust. United Acust. 95, 1032–1039. https://doi.org/10.3813/AAA.918235
Wei, W., Botteldooren, D., Van Renterghem, T., 2014. Monitoring sound exposure by real time measurement and dynamic noise map, in: Forum Acusticum 2014.
Wernet, G., Bauer, C., Steubing, B., Reinhard, J., Moreno-Ruiz, E., Weidema, B., 2016. The ecoinvent database version 3 (part I): overview and methodology. Int. J. Life Cycle Assess. 21, 1218–1230.
105
10 Annex I
Paper I
Meyer, R., Benetto, E., Igos, E., Lavandier, C., 2016. Analysis of the different techniques to include noise damage
in life cycle assessment. A case study for car tires. Int. J. Life Cycle Assess. 1–14.
Received: 3 June 2016 /Accepted: 23 August 2016# Springer-Verlag Berlin Heidelberg 2016
Abstract
Purpose Despite that different methods for the inclusion oftransport noise in life cycle assessment (LCA) have been pro-posed, none of them has become consensual. Leveraging acase study on car tires, this paper aims at comparing twoamong these characterization approaches to identify strengthsand weaknesses and to investigate the relative contribution ofnoise to human health (in disability-adjusted life years(DALYs)) as compared to other environmental stressors.Methods The case study analyzed two tires showing differentacoustical properties. The two methods applied are the onedeveloped by Müller-Wenk and further improved by otherauthors and the recent one proposed by Cucurachi. Thesetwo methods were adapted to the case study, and a full LCAstudy of the car tires was carried out. Both uncertainty andsensitivity analyses were performed.Results and discussion Both methods highlight the potentialhigh contribution of noise damage to the DALYs generated bycar tires, even considering parameters’ uncertainties. Thisstudy shows therefore the necessity to integrate noise impactin LCA in a broader way. Both methods present coherentresults regarding the environmental performance differencesbetween the two products. However, the absolute DALYscores differ by eight orders of magnitude, probably because
Cucurachi’s methods overestimate the damages. The analysisof modeling choices and parameter uncertainties could notexplain this difference.Conclusions Noise impact on human health has to be includ-ed in LCA, and additional efforts should focus on the charac-terization modeling since there is not yet a consensual methodfor a systematic integration. The case study shows that theimprovement of tire design can efficiently reduce noise impacton human health. Both methods have advantages and incon-veniences. We think that it is possible to elaborate a methodcombining the strengths of both approaches. An incrementalapproach used on accurate localized and temporalized dataprocessed with noise propagation software could providecharacterization factors for a set of archetypes. This shouldbe a good compromise for a method allowing systematic in-tegration of noise impact in LCA.
Life cycle assessment (LCA) is a standardized methodologyto evaluate the potential environmental impacts generated by aproduct or a system along its life cycle (ISO 2006). Mainfeatures are to avoid the displacement of pollution (from onelife cycle step to another) and to consider possible trade-offsamong impact categories (e.g., reducing climate change im-pacts while increasing eutrophication effects).
The World Health Organization (Fritschi et al. 2011) report-ed very comprehensively on the burden of disease from envi-ronmental noise, highlighting noise as a public health problem.As a result, the effects of noise should be reflected in the assess-ment of products involving significant transportation steps as
well as of large-scale systems like regional mobility. Despitethis need, the integration of traffic noise damages on humanhealth in LCA is still debated. Although several methodolog-ical attempts were advanced to fill in this gap, none of themhave become consensual so far. The first operational methodwas developed by (Müller-Wenk 2002, 2004) to evaluate thedisability-adjusted life years (DALYs) on human health due toadditional noise levels generated by cars and lorries, usingannoyance as midpoint. This method was then adapted byDoka (2003) who used the SonRoad model (Heutschi 2004)for the determination of emission levels. Later, Althaus et al.(2009) included data from SonRoad and from Steven (2005)to better distinguish different types of vehicles and situations.Nielsen and Laursen (2005) defined the noise nuisance impactpotential in person-seconds, but unfortunately, their modeldoes not allow repeatability because of the lack of necessaryinformation. The last development of the method developedby Müller-Wenk (2002, 2004) was achieved by Franco et al.(2010). Finally, Cucurachi et al. (2012) and Cucurachi andHeijungs (2014) developed a method to characterize any typeof noise (not only due to transportation), expressed in person-Pa*s (number of people exposed to a certain sound pressurefor a certain time duration). The latter two approaches havenever been applied on a case study and will be analyzed in thispaper. Concerning the application of noise characterization toLCA case studies, several examples related to roadtransportation are available in literature. PRé Consultants(2001) carried out an LCA of tires with data from nineEuropean manufacturers which integrated noise potential ef-fects but only qualitatively. Hofstetter and Müller-Wenk(2005) studied the possibility of monetization with five differ-ent approaches, showing a large variation of outcomes de-pending on the method chosen. Finally, Huijbregts et al.(2006) included traffic noise impacts into LCA of dwellingsby elaborating three possible traffic scenarios and calculatingthe resulting noise exposure, before linking it to DALYs usinga linear relationship between noise level in decibels (dB) andhuman health impact in DALYs.
The different methods still lead to significant variations inthe assessment of noise damage, thus making the contributionof noise to the overall environmental impacts of a functionalunit rather variable and finally very uncertain. As a conse-quence, it remains difficult to derive from the assessment con-crete improvement actions and opportunities for the productsconcerned and large-scale systems. The use of the DALYunitto assess noise damage to human health is multifold. First, itallows to compare methods with different midpoints, by scal-ing them to a common unit. Second, a midpoint approach canbe very practical but in order to compare the damage on hu-man health from different sources, the use of endpoint is man-datory. In the case of noise damage, we claim that stopping ata midpoint indicator will probably lead to neglecting of thisimpact category as it is often attributed to a lower importance
as compared to other impact categories like greenhouse effect.This does not entail that the endpoint level of characterizationshall be systematically preferred to the midpoint one for allimpact categories.
The objective of this study is threefold: (i) to analyze howthe different methods for noise damage assessment are sensi-tive to different input noise levels; (ii) to compare the resultsobtained from different methods; and (iii) to evaluate the im-portance of noise damage as compared to other environmentalcontributions on human health (in DALYs). The first part ofthe paper explains the different methods used to assess theimpact of noise; the second part briefly describes the conven-tional LCA of the tires. The results are then interpreted, com-paring the conclusions from different noise characterizationmethods. In a third part, the models will be analyzed usinguncertainty and sensitivity analyses. Finally, limits and furtherdevelopment of these methods will be addressed.
2 Critical review of noise assessment methods
The two methods investigated are the one developed byFranco et al. (2010), considered as the last improvement ofthe method from Müller-Wenk (1999, 2002, 2004) and Doka(2003), and the one developed by Cucurachi et al. (2012) andCucurachi and Heijungs (2014) who took a radically differentapproach. A case study on car tires is considered as a seminalexample of a functional unit to highlight and discuss the dif-ferences between methods. The necessary adaptations to eachstep of the cause-effect chain for the case of tire noise aredescribed in the following sections.
The functional unit is a tire used over 1 km. Two productsare analyzed: tire 1 and tire 2, from the same tire manufacturer,being summer tires with a size of 195/65R15 and a speedindex H. Significant efforts were made by the manufacturerto improve the characteristics of tire 2 (weight, oil content, androlling resistance). The changes in the tire structure and com-position lead to a significant decrease in fuel consumption, tirewear, and noise emission, as detailed later in Sects. 2.1 and3.2. The relative difference between the two tires will be stud-ied to estimate the potential environmental improvement, in-cluding noise impact.
Three types of roads are considered: urban roads, non-urban roads, and motorways. The average speed has beendefined for each road type. National traffic statistics from dif-ferent countries show important differences in traffic, roaddistribution, and population density, influencing noise levels.While the question of regionalization in noise impact assess-ment will be discussed further, this study focuses, as a firststep, on a single country, France, as done by Müller-Wenk(2002, 2004) for Switzerland, de Hollander et al. (1999) forthe Netherlands, and Franco et al. (2010) for Spain. The trafficis also split between three periods of the day: daytime,
Int J Life Cycle Assess
evening, and night. First, background noise is calculated usingpublicly available data on traffic and vehicle type repartitionand speed. Then, the increase of noise level due to an addi-tional tire is calculated and converted into damage on humanhealth. To calculate emission levels, both from backgroundand additional tire noise, the NMPB 2008 (Dutilleux et al.2010) was used and is detailed in Sect. 2.1.
2.1 NMPB 2008
The road noise prediction method NMPB 2008 is a state-of-the-art vehicle noise-emission model. It is a reference methodused under the European Directive 2002/49/EC (Directive2002) relating to the assessment and management of environ-mental noise.
NMPB 2008 gives different equations to calculate the noiseof light and heavy vehicles called type 1 and type 2. Theequivalent sound emission power (in dB) per meter of linesource for a flow rate of one vehicle per hour is noted Lw/m/
veh. It is the result of the rolling (Lr,w/m/veh) and motor (Lm,w/m/veh) components, calculated for different situations. For thiscase study, the sound level power and therefore Lw/m/veh needto be determined for the different types of roads and vehicles.
For tire 1 and tire 2 on vehicle type 1, experimentationsfrom standardized tests carried out by the manufacturer mea-sured sound pressure levels for a speed between 25 and120 km/h. The obtained data, LAmax, represents the maximumsound level at 7.5 m from the tire and 1.2 m from the ground,which is converted into Lr,w/m/veh based on Eq. 1 from NMPB2008:
Lr;w=m=veh ¼ LAmax−10*log vð Þ−4:4 ð1Þ
To extrapolate the results for the four tires of cars, an addi-tional +10*log(4) has been added to the Lr,w/m/veh. In NMPB2008, the rolling component for an aged surfacing categoryR2 (most common) with average tires is Lr,w/m/Type 1 = 55.4 +20.1*log(v/90) where v is the speed in kilometers per hour. Bydoing a linear regression between Lr,w/m/veh calculated for tire 1and tire 2 and log(v/90), the following equations are obtained(see Fig. 1): Lr,w/m/Tire 1 = 54.247 + 16.05*log(v/90) and Lr,w/m/Tire 2 = 49.316 + 11.363*log(v/90).
The difference between tire 1 and tire 2 is clearly visible inFig. 1, since tire 1 is similar to the reference rolling componentof NMPB 2008, for an aged surfacing category R2. It does notseem useful to add a reference tire in this case study. Theemission noise level from NMPB 2008 will be used for theexisting traffic noise, prior to the addition of tires 1 and 2.
Based on NMPB 2008, the motor component is calculated,assuming a steady speed, and added to the rolling componentto obtain the emission power level Lw/m/veh of type 1 vehicle,with tire 1, tire 2, and NMPB 2008 reference, and type 2vehicle based on NMPB 2008 for the three roads. An agedsurfacing category R2 was assumed, and the mean speed foreach road type is given in Sect. 4.1.
The emission level of the two tires needs now to be con-verted in DALYs for a small additional amount of tire 1 or tire2 added to the existing traffic.
2.2 Franco method
2.2.1 Assessment of the noise level increase
The method described here is adapted from Müller-Wenk(1999, 2002, 2004) and Franco et al. (2010). First, the trafficN (expressed in vehicle/hour) for vehicle type i on road type j(urban roads, non-urban roads, and motorways) during dayperiod k (day, evening or night) is determined with Eq. 2 (withTraffici, the total number of vehicle kilometers for vehicle typei by year, and Timek, the number of hours in period k).
N i; j;k veh=h½ � ¼Traffic i veh� km=year½ � � Traffic share j %½ � � Traffic share k %½ �
In Eq. 3, the emission power level per meter of line source,Lw/m/i,j calculated in Sect. 2.1, is broken up by road type j andperiod k based on the traffic Ni,j,k.
Lw=m;i; j;k ¼ Lw=m=i; j þ 10� log N i; j;k
� �
ð3Þ
The global equivalent emission power level Lw/m,j,k is de-duced from the equivalent noise level of vehicle types 1 and 2(Eq. 4).
Fig. 1 Sound level of the rolling component for the different tires
Int J Life Cycle Assess
The aim is now to evaluate the effect on this global soundemission power of additional vehicles equipped with the tiresstudied. An increase of 0.01 vehicle-kilometers from vehicletype 1 (adapted for tires 1 and 2) was considered, while trafficfor vehicles of type 2 stays identical. Only the rolling compo-nent Lr,w/m was added since the additional impact of the tirenoise is analyzed and not of the whole vehicle (excluding themotor component Lm,w/m). But even taking it into accountwould not have made a big difference because at the speedsused in this case study, the motor component is negligiblecompared to the rolling component. The rationale of settingsuch a marginal increase (0.01 vehicle-kilometers) is to assurelinearity between vehicle amount and sound power level. Theadditional sound power relative to the traffic increase is cal-culated in Eq. 5 and used to determine the equivalent soundemission power level L’w/m,j,k (as in Eq. 4).
The difference ΔLw/m,TireX,j,k between L’w/m,j,k and Lw/m,j,krepresents the marginal increase of noise level due to the ad-dition of 0.01 vehicle-kilometers of the rolling component oftype 1 vehicle equipped with four tires BTire X^ (X = {1,2})on each type of road j during each time period k. To obtain theimpact of 1 km driven, ΔLw/m,TireX,j,k is first divided by∆Vkmj,k, amount of vehicle-kilometers responsible for the ad-ditional 0.01 vehicle-kilometers of the rolling component oftype 1 vehicle for road j during period k (Eq. 6). The resultrepresents the increment in emission power level due to onevehicle-kilometer driven on road j during period k, which can
be distributed for the whole network and day period based ontraffic shares (δLw/m,TireX,j,k calculated with Eq. 7).
Franco et al. (2010) used Lden (day-evening-night equiva-lent level, based on Eq. 8) as the primary indicator for theevaluation of environmental noise.
Lden ¼ 10�log12
24�10
Lday10 þ
4
24� 10
Leveningþ5ð Þ10 þ
8
24� 10
Lnightþ10ð Þ10
!
ð8Þ
The factors represent period durations, with a poten-tial penalty (5 dB for evening and 10 dB for night) toreflect their importance to human perception. The dif-ference between sound emission power level Lw/m,k andLk (Eq. 8) for period k depends only on the geometryconsidered and is named A, for the sake of simplicity.
Calculating the Lden by road type j, Eq. 9 is obtained:
Lden; j ¼ 10� log12
24�10
Lwm; j;dayþA
10 þ4
24� 10
L wmj
;eveningþAþ5
� �
10 þ8
24� 10
Lwm; j;nightþAþ10ð Þ
10
0
B
@
1
C
A
¼ 10� log12
24�10
Lw=m; j;day10 þ
4
24� 10
Lw=mj;eveningþ5ð Þ10 þ
8
24� 10
Lw=m; j;nightþ10ð Þ10
!
þ A ð9Þ
L’den,j, the day-evening-night equivalent level obtained af-ter the addition of one vehicle-kilometer over the whole net-work and the whole day, is described in Eq. 10.
L0den;TireX; j ¼ 10� log12
24�10
Lw=m; j;dayþδLw=m;TireX; j;day10 þ
4
24� 10
Lw=mj;eveningþδLw=m;TireX; j;eveningþ5ð Þ10 þ
8
24� 10
Lw=m; j;nightþδLw=m;TireX; j;nightþ10ð Þ10
!
þ A ð10Þ
Subtracting Lden,j to L’den,TireX,j, the increment in day-evening-night equivalent level on road type j due to the
addition of one vehicle-kilometer over the whole networkand the whole day is obtained. Doing this, one can then
Int J Life Cycle Assess
observe that the term A has absolutely no importance since thesubtraction cancels it. The last step to obtain the increment inday-evening-night equivalent levels due to one vehicle-kilometer over the whole network and the whole day is asummation on j as shown in Eq. 11. Since the normalizationhas been done before, this is basically an averaging of ∆Lden,jover the different type of roads respecting the traffic sharebetween these roads.
δLden;TireX ¼ ∑j
L0den;TireX; j−Lden; j� �
ð11Þ
2.2.2 Assessment of the additional number of highly annoyed
people
The assessment of the additional number of highly annoyedpeople is one of the improvements of Franco et al. (2010)compared to Müller-Wenk (1999, 2002, 2004) and Doka(2003). It is based on the work of Miedema and Oudshoorn(2001) giving the polynomial approximations of the dose re-sponse curves relating Lden levels at the most exposed façadeof dwellings and self-reported annoyance. The exposure of thepopulation in Lden has been extracted from the strategic noisemap, calculated by different cities under the EuropeanDirective 2002/49/EC, and given in a number of peopleNp(g) exposed to noise level in decibel range g of 5 dB.
Franco et al. (2010) differentiated the polynomial approx-imation from Miedema and Oudshoorn (2001) for the centerof each decibel range. The result ∆%HA/∆Lden is the addition-al percentage of highly annoyed people by additional numberof decibels around this point. Doing this for each decibel rangeg, the additional number of highly annoyed people is deter-mined (Eq. 12).
ΔHATireX ¼ ∑g
N p gð Þ �Δ%HA=ΔLden gð Þ � δLden;TireX ð12Þ
Franco et al. (2010) considered the additional number ofhighly annoyed people to be a good midpoint for the integra-tion of noise in LCA. However, to compare this method withthe one developed in Cucurachi et al. (2012) and Cucurachiand Heijungs (2014), this result needs to be converted inDALYs.
2.2.3 Assessment of the DALYs per additional highly annoyed
people
The World Health Organization (Fritschi et al. 2011) made areview of the studies linking health impact and annoyance(highly annoyed people calculated from Lden level) and rec-ommended a disability weight (DW) factor of 0.02 DALY/highly annoyed people (uncertainty range 0.01–0.12), whichis therefore used in this study.
2.3 Cucurachi method
The methodology translates sound powers (in Joules) into thenumber of people that are exposed to a certain sound pressurefor a certain period of time (in person-Pa*s). The characteri-zation factors (in Person-Pa*W) related to elementary flows(in J) are calculated for archetypal situations depending ontheir frequency (eight octave bands), their time period (day,evening, and night), and their place (urban, suburban, rural,industrial, indoor). Cucurachi and Heijungs (2014) also pro-posed an Excel sheet to develop user-defined characterizationfactors, which could not be used due to the lack of availableinput parameters (e.g., temperature, relative humidity, distancefrom source to receiver, background noise). Regarding thearchetypal characterization factors, the measurements of thetire manufacturer are given in dB(A) (dB values aggregatedand weighted across frequency range) and only factors forunspecified frequency (calculated with central frequency of1000 Hz) could be applied.
The NMPB 2008 methodology determines the sound pow-er level of a source Lw from the emission power level permeter of line source for a flow rate of 1 vehicle per hourLw/m/veh using Eq. 13.
Lw;TireX ¼ Lr;w=m=TireX þ 10*log10 vð Þ þ 30 ð13Þ
Based on the speed and the Lw/m on the different roads, thesound power level of the rolling component (due to tires) canbe calculated for each road. The sound power level differen-tiated per type of road j in dB(A) can be converted into soundpower in Watts via Eq. 14 where Wref = 1 pW (ISO 9613-2,1996).
W TireX; j ¼ W ref � 10Lw;TireX; j=10 ð14Þ
To express sound power valuesWj for the functional unit inJoules, one vehicle-kilometer is split between the differentroad types and time thanks to the traffic share values, and thendivided by the mean speed. The obtained time Tj,k representsthe time one vehicle should spend on each road type j duringeach period k to create this additional one vehicle-kilometer inthe system. In other words, Eq. 15 represents the time ofemission of the corresponding source.
The sound energy Ej,k (in Joules), product of Tj,k andWj, ismultiplied by the corresponding characterization factor de-fined by Cucurachi and Heijungs (2014), depending on geo-graphical (rural conditions considered for motorways andnon-urban paths) and temporal contexts (day period).
Int J Life Cycle Assess
The classification of the roads defined in this study doesnot fit the predefined situations from Cucurachi (e.g., Bnon-urban road^ includes suburban and rural locations); this willbe discussed later. The differentiated impacts obtained (in per-son-Pa*s) are summed up at the midpoint. In the last step, theimpacts obtained in person-Pa*s are converted to DALYswithone of the two mid_to_end conversion factors proposed byCucurachi and Heijungs (2014) based on de Hollander et al.(1999) and the WHO study (Fritschi et al. 2011). As stated bythe authors, these factors may be used to provide a measure ofthe noise impacts at the endpoint level. However, the authorsalso warned against uncertainties due to the assumption oflinearity. Comparing the results obtained from these factorswith the ones calculated with another method can be a goodway to assess the reliability of the factors.
3 LCA of tires without noise consideration
3.1 Goal and scope definition
Since the aim of the study is also to evaluate the contributionof noise to human health damage as compared to other envi-ronmental burdens, a comprehensive LCAwas carried out fortire 1 and tire 2. The LCA included the production of rawmaterials, car tire production, retailing, use of the tire, andseveral end-of-life pathways, for a functional unit defined asthe use of one tire over 1 km. The corresponding inventorydata is described in the next paragraphs.
3.2 Life cycle inventory
Inventory data for tire 1 was taken from PRé Consultants(2001) and further updated to reflect the current backgroundprocesses and use situations (e.g., share of end-of-life routes).For tire 2, representative data for the year 2010 was collectedfrom the tire manufacturer. In order to express the inventoryflows according to the functional unit, an average mileageduring the lifetime of the tire is considered, respectively40,000 and 46,000 km for tire 1 and tire 2.
3.2.1 Raw materials and tire production
The consumption and transport of raw materials were basedon the composition of the tires. Some simplifications werenecessary because of the lack of inventory data for a fewraw materials and for confidentiality reasons. Firstly, all thetypes of synthetic rubber were merged into only one category,as well as for accelerators, anti-degradant, textile, and wires.The type of resin was unspecified, and therefore, phenolicresin was considered as a proxy. Finally, no specific data couldbe gathered for the retarder, adhesion promoter, co-salt, andpeptizer, which were therefore excluded from the life cycle
inventory (LCI). These materials represent less than0.16 % of the tire mass, and therefore, their contributionto the overall environmental impact of the tire is likelyto be negligible. For the production of tires, energyconsumption (electricity and natural gas) and relatedemissions were estimated according to data from the tiremanufacturer, completed by information from ecoinvent2.2 (Frischknecht et al. 2005; PRé Consultants 2001).
3.2.2 Use phase
Two parameters were taken into account for the use phase: tiredebris emissions and fuel consumption. The total amount oftire debris was retrieved from PRé Consultants (2001) and isequal to 30 mg/km. The composition of tire debris was basedon measurements by the tire manufacturer and from literaturedata (ETRMA 2010; Ntziachristos and Boulter 2009; Null1999; PRé Consultants 2001; Rauterberg-Wulff 2003; TenBroeke et al. 2008), including particulate matter (with a diam-eter equal or lower than 10 and 2.5 μm), polycyclic aromatichydrocarbons, distillate aromatic extract (treated or not), andoil and zinc emissions.
To calculate the fuel consumption linked to the tire, theaverage car fuel consumption was estimated and allocated tothe tire. From PRé Consultants (2001), the average consump-tion for a car is 7.5 L/100 km with a share of 50/50 betweengasoline and diesel cars. Considering the mean sensitivityfriction coefficient, the authors estimated a consumption of1.2 L/100 km for the four tires. This data was taken intoaccount for tire 1 while the consumption of tire 2 was basedon the gain of rolling resistance (7.8 kg/kg instead of 10.5 kg/kg for the tire 1). Based on literature (ChemRisk 2009; Evanset al. 2009; Krzyżanowski et al. 2005; Weissman et al. 2003),0.26 % of fuel reduction is obtained per 1 % of rolling resis-tance reduction. The fuel consumption is therefore 0.7 L/100 km for four tires in the case of tire 2.
3.2.3 End-of-life of tires
The end-of-life route data was provided by the tire manufac-turer and taken into account retreading (8 %), direct reuse/export (10 %), material recovery (40 %), energy recoveryvia incineration in cement kilns and power plants (38 %),and landfill (4 %).
3.3 Life cycle impact assessment
The ReCiPe (Goedkoop et al. 2009) method, following thehierarchist perspective and average weighting set (H, A),was chosen for the assessment at endpoint level (impacts onhuman health, ecosystems, and resources expressed inDALYs, species, year and dollars, respectively). The integra-tion of noise impacts in DALYs is therefore possible and can
Int J Life Cycle Assess
be compared with other impacts on human health such asclimate change, ozone depletion, or human toxicity.
4 Evaluation of the reliability of the noise impact
The road traffic was obtained by averaging several sourcesand time series from the UNECE database (Lacour andJoumard 2002) and the French Road Safety organization(French Road Safety reports, Observatoire NationalInterministériel de la Sécurité Routière (ONISR) [WWWDocument] (n.d.): 430 billions of vehicle-kilometers for cars(type 1) and 29 billions of vehicle-kilometers for trucks.
Finding relevant data for the traffic distribution among thetime periods is not trivial. The data listed in Table 2 is issuedfrom a vast area around Lyon (France) and were provided byAcoucité (BObservatoire de l’environnement sonore de laMétropole de Lyon [WWW Document], n.d.^) who realizedthe noise maps for the Lyon area.
The corresponding area represents 2.1 and 2.4 % of thetotal French traffic for vehicle types 1 and 2, respectively,and covers different situations (from urban to rural). It canbe therefore considered as a good representation for the wholecountry. Also, these values are similar to ones from Müller-Wenk (2002, 2004), i.e., 86.9 % during daytime and 13.1 %during nighttime.
French population equals to 64.5 million individuals at thebeginning of 2016 according to the French National Institute
of Statistics and Economic Studies (Institut national de lastatistique et des études économiques, INSEE [WWWDocument], n.d.). To apply the Franco method, the noise ex-posure of the population is also needed. No specific data isavailable covering the whole French population. However, theEuropeanDirective 2002/49/EC requires cities withmore than100,000 inhabitants to elaborate noise maps and calculate theexposure to noise of their population. The data has been col-lected for 2007 and 2012 (Noise Observation and InformationService for Europe, NOISE [WWWDocument], n.d., NOISEplatform, available at http://noise.eionet.europa.eu/). In thecase of France, the dataset for 2007 is more complete andtherefore used in this case study (Table 3).
The population exposed to more than 75 dB is excludedfrom the case study since the polynomial approximations fromMiedema and Oudshoorn (2001) are not verified after 75 dB.However, this limitation concerns only a very small percent-age of the population. The population under 55 dB is consid-ered as non-affected. It would have been better to know thepercentage of population in the 50–55-dB range, but there isno data available for this range. The population covered bythis study is around 22 million, i.e., 34 % of the whole popu-lation considered. The availability of accurate data for onlyone third of the considered population is a significant short-coming of the study. Moreover, data are concerned mostlywith cities with a population of more than 100,000 inhabi-tants, and one can expect a different exposure for smaller citiesand rural area.
The number of additional people highly annoyed by oneadditional decibel has been calculated for each middledecibel-range of exposure (Table 4). These coefficients aredifferent from the one of Franco et al. (2010). Despite theauthors’ claim, the factors used do not come from the polyno-mial approximation of the synthesis annoyance curves extract-ed from Miedema and Oudshoorn (2001) for percentage ofhighly annoyed people but from the one for lowly annoyedpeople. As a result, different coefficients are obtained which
Table 1 Traffic data inventoried
Roadtype
Length(km)
Average speedtype 1 (km/h)
Average speedtype 2 (km/h)
Percentage oftraffic (%)
Motorway 11,243 115 90 23
Non-urbanroad
387,018 80 75 47
Urbanroad
630,000 50 50 30
Table 2 Traffic distribution among the different periods of the day
Time period Percentage of type 1traffic (%)
Percentage of type 2traffic (%)
Day 72 70
Evening 21 15
Night 7 15
Table 3 French population’s exposure coming from the NOISEplatform
dB range 55–59 60–64 65–69 70–74 >75
Percentage of the population 22.9 % 21.3 % 15.7 % 9.0 % 1.6 %
increase with the decibel-range instead of being relativelyconstant in the case of the equation for lowly annoyed people.
The collection of data shows that assessing the noise im-pact on a whole country leads to significant approximations,due to lack of data but also to the modeling approach. Forexample, assuming an average speed for a whole road net-work or uniform traffic over a whole country is not represen-tative of the reality. Noise varies quickly in space and time andcan only be assessed accurately over a specific location and ona short time scale. Studying noise impact over a large territoryforces the practitioner to make unrealistic assumptions, andthe result can only give a very approximate picture of reality.In order to achieve more accurate results, a smaller scale needsto be focused upon. Section 4.2 focuses on data uncertainty,whereas the pertinence of the approximations due to the modelitself is not considered.
4.2 Uncertainty distributions of models’ parameters
Uncertainty distributions were defined for all the parametersof the models. For most of them, only a few values wereavailable and therefore normal distributions were assumed,using the mean and standard deviation of the values to setthe distributions.
The mean and standard deviation of the Total number
of vehicle-km of type 1 and the Total number of vehicle-kmof type 2 were derived from UNECE database (Lacourand Joumard 2002) and the French Road Safety organiza-tion (French Road Safety reports, Observatoire NationalInterministériel de la Sécurité Routière (ONISR) [WWWDocument] (n.d.). For the length of the three types ofroad, it is considered as a reliable value, assuming a stan-dard deviation of 5 %. For the different average speeds,the distributions have been estimated with yearly datafrom the French national road safety website (ONISR).For the repartition of the traffic between the differenttypes of roads, data from Lacour and Joumard (2002)were taken (similar data for motorways found in aFrench national road safety website). The resulting distri-bution of traffic between road types based on thesesources is very different from the one considered byMüller-Wenk (2004), mostly due to national differencesin both road network and usage. To model uncertaintieson percentage data, a Dirichlet distribution was used (toensure a sum of 1), with assumed standard deviation
around 10 %. For the percentage of traffic between thedifferent periods of the day, the data used is consideredrepresentative of the French territory, therefore assuminga standard deviation around 4 %.
For the emission power level of tire 1 and tire 2, two typesof uncertainty have been defined. The first uncertainty, calleddBNMPB, comes from NMPB 2008, which gives a 95 % con-fidence interval of ±2.5 dB(A) for the rolling component, witha distribution similar to the Gaussian form. Therefore, a nor-mal distribution with a standard deviation of ±1.25 dB(A) waschosen for all the type 1 vehicles. NMPB 2008 gives a 95 %confidence interval of ±3.2 dB(A) for the rolling componentof type 2 vehicles. Calculating independently the uncertaintyof type 2 vehicles would not be correct, so the uncertainty fortype 2 is equal to dBNMPB/2.5*3.2.
The second source of uncertainty for the emission powerlevel of tire 1 and tire 2, called dBexp, comes from the mea-surements. dBexp has been defined as the minimum level ofdecibel required to envelop every measurement points withinLr,w/m/Tire x ± dBexp,Tire x. This error is dependent on the tire:±1.3 dB for tire 1 and ±1.8 dB for tire 2. These values havebeen set as the standard deviation of a normal distributioncentered on 0.
A standard deviation of 2 % was set for the populationnumber to represent the good quality of data from INSEE.
For the exposure of the population to Lden, extrapolation ofdata covering only a third of the total population can lead toimportant mistakes. To avoid additional hypothesis, the max-imum value for each decibel-range is the one from the NOISEplatform (big cities most exposed to environmental noise) andis multiplied by the percentage of concerned population toobtain the minimum value (i.e., no exposition for the popula-tion not covered by the NOISE platform). A uniform uncer-tainty distribution was set between this minimum and maxi-mum for the exposure in each decibel-range.
Concerning the number of additional people highlyannoyed by the additional decibels of Lden, uncertainties comefrom the exact position in a given decibel range because thiscoefficient is lower in the lower part of a given decibel rangethan in the upper part. A uniform distribution is defined basedon coefficients in both extremities.
The weighting factor linking additional highly annoyedpeople to DALYs have been modeled from Fritschi et al.(2011) values. To fit the median given by the WHO, twotriangular distributions have been used, one between the min-imum and the median and the other between the median andthe maximum. Then, when using these distributions, for asample of size N, N/2 elements are coming from eachdistribution.
Regarding the characterization factors for theCucurachi model, the author suggested an uncertainty oftwo orders of magnitude above and below the valuesspecified in Cucurachi and Heijungs (2014). A triangular
Table 4 Additional percentage of highly annoyed people by Lden at agiven exposure
dB 55 57.5 60 62.5 65 67.5 70
Δ%HA/ΔLden 0.64 0.78 0.95 1.17 1.42 1.71 2.03
Int J Life Cycle Assess
distribution has been chosen with a logarithmic base of10, where the median equals the characterization factorsby Cucurachi. As only two values for the Mid_to_end
factor are available, a uniform distribution has been setbetween these two values (as conservative hypothesis).
The different parameters, their mean, standard deviation,and distribution are given in Tables 5 and 6.
4.3 Sensitivity and uncertainty analysis methods
Sensitivity analysis is a recommended step by the ISO standards14040-14044 (ISO 2006) to observe the influence of the assump-tions and choices made during the LCI and LCIA phases. Withthe set of uncertainties described above, uncertainty and sensitiv-ity analyses have been made using the R software.
Table 5 Parameters and their uncertainty distribution for the Franco model
Franco model: parameters and their uncertainties
Parameter Mean(Baseline)
Standarddeviation
Distribution used
Total number of vehicle-km of type 1 4.30E+11 3.25E+10 Normal (429580500000., 32537748003.2381)
Total number of vehicle-km of type 2 2.90E+10 7.11E+9 Normal (28990750000., 7109647384.26396)
Motorways’ length 11,243 562 Normal (11243., 562.)
Non-urban roads’ length 387,018 19,351 Normal (387018., 19351.)
Urban roads’ length 630,000 31,500 Normal (630000., 31500.)
Motorways’ average speed of type 1 115 15 Normal (115., 15.)
Non-urban roads’ average speed of type 1 80 10 Normal (80., 10.)
Urban roads’ average speed of type 1 50 5 Normal (50., 5.)
Motorways’ average speed of type 2 90 8 Normal (90., 8.)
Non-urban roads’ average speed of type 2 75 5 Normal (75., 5.)
Urban roads’ average speed of type 2 50 5 Normal (50., 5.)
Percentage of traffic on motorways 23.0 % 9.2 % Dirichlet (4.6, 9.4, 6.0)
Percentage of traffic on non-urban roads 47.0 % 10.9 % Dirichlet (4.6, 9.4, 6.0)
Percentage of traffic on urban roads 30.0 % 10.0 % Dirichlet (4.6, 9.4, 6.0)
Percentage of traffic during the day of type 1 72.0 % 4.5 % Dirichlet (72, 21, 7)
Percentage of traffic during the evening of type 1 21.0 % 4.1 % Dirichlet (72, 21, 7)
Percentage of traffic during the night of type 1 7.0 % 2.5 % Dirichlet (72, 21, 7)
Percentage of traffic during the day of type 2 70.0 % 4.6 % Dirichlet (70, 15, 15)
Percentage of traffic during the evening of type 2 15.0 % 3.6 % Dirichlet(70, 15, 15)
Percentage of traffic during the night of type 2 15.0 % 3.6 % Dirichlet (70, 15, 15)
dBNMPB 0 1.25 Normal (0, 1.25)
dBexp,Tire 1 0 1.30 Normal (0, 1.3)
dBexp,Tire 2 0 1.80 Normal (0, 1.8)
Population 6.45E+7 1.29E+6 Normal (64500000, 1290000)
Percentage of exposed population 55–60 dB 22.90 % Uniform (0.229/3, 0.229)
Percentage of exposed population 60–65 dB 21.30 % Uniform (0.213/3, 0.213)
Percentage of exposed population 65–70 dB 15.70 % Uniform (0.157/3, 0.157)
Percentage of exposed population 70–75 dB 9.00 % Uniform(0.090/3, 0.090)
Additional percentage of highly annoyed per additional Lden55–60 dB
0.81 % Uniform (0.00639, 0.00954)
Additional percentage of highly annoyed per additional Lden60–65 dB
1.21 % Uniform (0.00954, 0.01418)
Additional percentage of highly annoyed per additional Lden65–70 dB
1.77 % Uniform (0.01418, .02029)
Additional percentage of highly annoyed per additional Lden70–75 dB
2.47 % Uniform (0.02029, .02789)
Weighting factor converting highly annoyed in DALYs 0.02 1/2 Triangular (0.01, 0.02, 0.02) 1/2 Triangular (0.02,0.02, 0.12)
In bold: target parameter in the multivariate distribution
Int J Life Cycle Assess
For the uncertainty analyses, a Monte Carlo approachwith two samples of 100,000 iterations was adopted.Regarding global sensitivity analysis, Sobol indices ofeach parameter were calculated using the soboljansen
function from the R package sensitivity, the samplesgenerated with Monte Carlo approach, 500 bootstrapreplicates, and a level for bootstrap confidence intervalsof 0.95. The first-order Sobol index Si represents thecontribution to the output variance of the main effectof parameter i, standardized by the total variance toprovide a fractional contribution. The total-effect Sobolindex STi measures the contribution to the output vari-ance, including all variance caused by its interactions,of any order, with any other input variables. The differ-ence between these two indices provides information onthe level of interaction in the system.
5 Results interpretation
5.1 Integration with other impact categories
When integrating noise impacts into the overall life cy-cle impacts of the tire, its contribution is 8.1 % for tire1 and 76.7 % for tire 2 with the Franco model. Thiscan be a surprising result because tire 1 produces morenoise than tire 2. However, other sources of DALYshave been considerably reduced between the two tires,in a bigger proportion than noise. This is because, de-spite the decrease in absolute value, the importance ofnoise between the different sources of DALYs has in-creased in tire 2 compared with tire 1. In the case ofthe Cucurachi model, the impact of noise is so high thatit accounts for almost 100 % for both tires.
Table 6 Parameters and theiruncertainties’ distribution for theCucurachi model
Cucurachi model: parameters and their uncertainties
Parameter Mean(baseline)
Standarddeviation
Distribution used
Motorways’ average speed of type 1 115 15 Normal (115., 15.)Non-urban roads’ average speed of
type 180 10 Normal (80., 10.)
Urban roads’ average speed of type 1 50 5 Normal (50., 5.)Percentage of traffic on motorways 23.0 % 9.2 % Dirichlet (4.6, 9.4, 6.0)Percentage of traffic on non-urban
roads47.0 % 10.9 % Dirichlet (4.6, 9.4, 6.0)
Percentage of traffic on urban roads 30.0 % 10.0 % Dirichlet (4.6, 9.4, 6.0)Percentage of traffic during the day of
type 172.0 % 4.5 % Dirichlet (72, 21, 7)
Percentage of traffic during theevening of type 1
21.0 % 4.1 % Dirichlet (72, 21, 7)
Percentage of traffic during the nightof type 1
7.0 % 2.5 % Dirichlet (72, 21, 7)
dB (NMPB) 0 1.25 Normal (0, 1.25)dB (exp ; tire 1) 0 1.30 Normal (0, 1.3)dB (exp ; tire 2) 0 1.80 Normal (0, 1.8)Characterization factor, motorways,
logbase = 10)Mid to end factor 0.000213 Uniform (0.000029061, 0.000213)
In bold: target parameter in the multivariate distribution
Int J Life Cycle Assess
In most of the LCA studies, the impact of noise is not takeninto account. The results obtained in our case study show thatthe impact of noise can be the most important among thedifferent impacts on human health caused by a tire. This resultinvokes that more effort is needed toward the systematic inte-gration of noise into LCA. It however also raises many ques-tions. Are the noise assessment models representativeenough? Are they robust? Can they be improved? In theSect. 5.2, we will analyze in more detail the two studies andtry to shed light on weaknesses of the existing model.
5.2 Relative change
The first aspect when analyzing a noise impact assessmentmethod is how it reacts to noise differences in the inputs. Tocalculate the impact between the two tires on human health,the relative change calculated as in Eq. 16 must be analyzed.
Relative change ¼DALYTire 1−DALYTire 2
max DALYTire 1;DALYTire 2ð Þð16Þ
Using the baseline values, the Franco method gives arelative change of 64.8 % between the two tires. TheMonte Carlo sampling with 200,000 iterations gives, forthe relative change, a mean of 60.0 %, a median of64.7 %, and a standard deviation of 21.3 %. The relativechange in the Franco method seems reliable. This methodvalidates the improvement between tire 1 and tire 2, eventaking into account the important uncertainties, if they havebeen modeled appropriately.
The sensitivity analysis on the Franco method shows thattwo parameters, dBexp,Tire1 and dBexp,Tire2, account for 84 % ofthe sum of the first-order Sobol indices and 99.6 % of the sumof total-order Sobol indices. It means that the uncertainties ofthe noise emission levels of the two tires are responsible for avery large part of the uncertainty about the relative change inDALYs between tire 1 and tire 2. More precise measurementsof the noise emission levels of the different tires coupled withwell-defined and narrower uncertainty and better fitting ofmeasurement curves will lead to smaller uncertainty on therelative change. The difference between the Sobol indices ofdBexp,Tire1 (S = 0.33, ST = 0.36) and dBexp,Tire2 (S = 0.63,ST = 0.67) is probably due to the difference in the standarddeviation of these two parameters. The difference between thefirst-order and total-order Sobol indices indicates that there areonly a few interactions in this model.
Regarding the Cucurachi method, the baseline valuegives a relative change of 56.0 % between the two tires,while the uncertainty analysis leads to a mean of 55.4 %,a median of 63.0 %, and a standard deviation of 32.2 %.This method also validates the improvement between tire
1 and tire 2, even if it is slightly lower than that with theFranco method.
Based on the total-order Sobol indices for the Cucurachimethod, dBexp for tire 1 and tire 2, explaining 46.9 % of theoutput variance, and the characterization factors, 50.9 % ofoutput variance, are responsible for most of the result uncer-tainty. As for the Francomethod, the contribution of dBexp,Tire1is approximately half of the contribution of dBexp,Tire2.Looking into detail into the characterization factors, the onesfor urban roads have a higher contribution, probably becauseof the larger part of the population impacted. Concerning thetemporal period, day characterization factors are higher thanthose during evening and night, probably because most of thetraffic occurred during the day. Characterization factors have alarge contribution on the uncertainty of the relative changebecause they have wide uncertainty distribution, roughly de-termined from suggestion from Cucurachi. Developers ofcharacterization factors should provide detailed uncertaintydistributions in order to allow proper uncertainties and sensi-tivity analyses.
The two methods do not show big differences whenassessing the relative change between tire 1 and tire 2. In bothcases, tire 2 is better and the reduction in DALYs issignificant.
5.3 DALY score
To analyze the score in DALYs resulting from the twomethods, it has been chosen to focus on tire 1 as analyzingtire 2 would naturally give similar results.
The obtained values were compared to the ones of Müller-Wenk (2002) and Franco et al. (2010) to check their validity.The level increase for 1000 vehicle-kilometers was 5.00e-7dBA (measured as LAeq) for Müller-Wenk (2002) and 3.80e-7 dBA in Lden for Franco et al. (2010), compared to 1.64e-8dBAwith the base values of this study. This lower value canbe explained by the longer road network (1,028,261 km in-stead of 13,872 km in Franco et al. 2010). On the other hand,the traffic on non-urban and urban roads is lower, so theresulting background noise is lower, which explains the small-er difference in Lden than in road length. Concerning the an-noyance, Müller-Wenk (2002) found 3.8e-2 cases of commu-nication disturbance and Franco et al. (2010) found 1.3e-3additional cases of annoyance. This study identified 1.0e-2additional cases of annoyance for tire 1. Differences can beexplained by the lower Lden increase, counterbalanced by apopulation 20 times bigger. Despite the differences betweenthe models used, the results obtained here are in the sameorder of magnitude than the ones from these two studies. Itdoes not seem to have major errors in the adaptation andapplication of the chosen methodology.
In the case of the Franco method, a value of 5.00e-8DALYs has been found for 1 vehicle-kilometer for tire 1 using
Int J Life Cycle Assess
the base values. A Monte Carlo run with 200,000 iterationsgives a median of 3.90e-8 DALYs, a mean of 5.82e-8 DALYs,and a standard deviation of 4.89e-8 DALYs. Uncertainties donot radically change the mean value but show an importantstandard deviation. Based on the sensitivity analysis onMonteCarlo samples, two parameters, dBexp,Tire1 and disability
weight, account for 84.0 % of the sum of the first-orderSobol indices and 99.6 % of the sum of the total-order Sobolindices. The first one, dBexp,Tire1, was already discussed be-fore. The second one, disability weight, is based on theFritschi et al. (2011) with S = 0.71 and ST = 0.81. In the caseof the Franco method, to achieve a better accuracy of theresult, it seems that the focus has to be put on diminishinguncertainties to link annoyance and human health. This resultis coherent with the Fritschi et al. (2011) presenting this link asvery uncertain due to the intrinsic difficulty of assessing noiseimpact and shows the importance of refining it to improveassessment reliability.
In the case of the Cucurachi method, a value of 5.85DALYs (based on the mid_to_end factor calculated byCucurachi from the WHO study) has been found for 1vehicle-kilometer for tire 1 using the base values. A MonteCarlo run with 200,000 iterations gives a median of 9.74DALYs, a mean of 16.95 DALYs, and a standard deviationof 22.77 DALYs. The difference between the median and themean is probably coming from the form of uncertainties cho-sen for the characterization factors.
The Cucurachi method has never been applied, to ourbest knowledge, to any case study. This means that thereis no other point of reference to check the result validity.However, we will compare the values coming from thetwo different methods summarized in Table 7. Takingthe mean, the Franco method gives 5.82e-8 DALYs andthe Cucurachi method gives 16.95 DALYs. There aremore than eight orders of magnitudes between the twomethods! This is a huge difference even for LCA. Evenif noise impact can be seen as difficult to assess and mea-sure, it is improbable that the damage on human healthdue to 1 vehicle-kilometer driven by one tire is anythingnear unity. Between these two results, the one fromFranco seems more realistic and reliable. Moreover, theconsideration of uncertainties in the case of the Cucurachimethod leads to a range of possible results whose averageis increased by a factor 3 compared to the case of basevalues. This means that the uncertainties are not
responsible for the very significant difference observedon the results obtained from the two methods.
For the Cucurachi method, the sensitivity analysesshow a significant difference between the sum of thefirst-order Sobol indices 0.72 and the sum of the total-order Sobol indices 1.42. It indicates strong interactionsin the model. Looking at the total-order Sobol indices,uncertainties on the input parameters dBexp,Tire1 explains9.3 % of the output variance and uncertainties about theemission dBNMPB explains an additional 8.6 %. Thehighest contribution comes from the characterization fac-tors accounting for 53.0 % of the output variance (higherfor urban roads and daytime). Impact on the uncertaintyof one characterization factor among others is directlylinked to the contribution of this road/period into thewhole DALY score. The contribution to the output varianceof themid_to_end factor is 17.7 % despite a very conservativeassumption of a uniform distribution between the two valuesgiven in Cucurachi and Heijungs (2014). Diminishing theuncertainties on the output implies refining the characteriza-tion factors and the link between the midpoint chosen byCucurachi, person-Pa*s, and the endpoint in DALYs.However, taking into account the really high result inDALYs, there are probably other problems to treat beforelooking at the uncertainty of the output.
6 Discussion
The impact of an additional amount of sound power is verydependent on the geographical situation because it depends onthe affected population (thus density), the existing sound lev-el, the topography, the absorption characteristics of the envi-ronment, etc. A lot of different parameters play a role in theimpact of an additional noise; consequently, it can be difficultto elaborate generic characterization factors. However, a re-gionalization of the life cycle impact assessment is not veryuseful if there is not a similar regionalization of the life cycleinventory. In the methods applied here, the inventory is doneat the national scale and it does not fit the predefined arche-types from Cucurachi and Heijungs (2014). The motorwaysand non-urban roads of our study can be used with the arche-types of unspecified location (chosen for this study), of rurallocation, or of suburban location. To evaluate the impact of
Table 7 DALY score and uncertainty for the two methods
Base Median Mean Standard deviation
Franco 5.00E−08 3.90E−08 5.82E−08 4.89E−08
Cucurachi 5.85 9.74 16.95 22.77
Table 8 Different location used for the characterization factors
Cucurachi Unspecified Rural Suburban
Tire 1 5.85 0.63 4.23
Tire 2 2.57 0.36 2.20
Relative change 0.56 0.42 0.48
Int J Life Cycle Assess
this choice, we calculated the DALY score and relative changewith three different possibilities for the location of motorwaysand non-urban roads. Table 8 shows the importance of thischoice.
Our incapacity to fit the inventory into the predefined ar-chetypes does not explain the high result coming from theCucurachi method. Looking at Cucurachi and Heijungs(2014), it is possible that these mid_to_end factors have beenoverestimated. To calculate it, the L95 value, noise level (dB)exceeding 95 % of time, was used coming from the fullBANOERAC report (EASA 2009). Even without lookingmore into details into the methodology, it is understandablethat a noise level exceeding 95 % of the time can be verydifferent from the equivalent noise level of the same area(equivalent noise level often chosen as a measure of the envi-ronmental noise). It means that Cucurachi and Heijungs(2014) may have underestimated the background noise andthus the calculation of the midpoint for the studied zone.Consequently, the mid_to_end factors can be largelyoverestimated.
Moreover, Cucurachi and Heijungs (2014) calculate themid_to_end factors in two studies accounting for the totalimpact on noise damage on human health. This approach doesnot take into account the possible non-linearity of the problemas already stated by the authors. In order to have a betterestimation of the mid_to_end factors, we advise to use anincremental approach. LCA considers, in most cases, smallvariations around a base scenario assumed to be representativeof the current conditions. An incremental approach wouldgive a mid_to_end factor closer to the modeled situation andthus would allow decreasing the uncertainty from non-linearity.
Regarding the general philosophy of these two differentmethodologies, they have both their pros and cons. TheFranco method used an incremental approach to assess theadditional impact of noise of a small increment in a specificsituation. This methodology has been developed by severalauthors and seems to give reliable results. However, it needsa specific modeling numerous data on the considered situa-tion. This can be time-consuming and would not allow sys-tematic integration of noise impact in LCA. Also, it wouldonly be possible for the foreground system.
On the other hand, the advantage of the Cucurachi methodto use archetypal situations and precalculated characterizationfactors can lead to a quick and systematic integration of noisedamage on human health in both foreground and backgroundof LCA studies. Of course, this exercise remains very difficult,considering that the characterization factors should show suf-ficiently low variability within an archetype and should beeasily understandable and applicable. However, theCucurachi method failed to give a plausible result and it isquite difficult to know why and how to improve thismethodology.
We think that there may exist a way to conciliate these twodifferent approaches: use the first approach to generate char-acterization factors in a given number of archetypal situations.This third way may be a good compromise, using the strengthof the two studied methods. The best way to study the feasi-bility of such characterization factors and to calculate themseems to use up-to-date data and tools. The EuropeanDirective 2002/49/EC forces European cities of more than100,000 inhabitants to generate noise maps and to collectthe needed data. This data could be used with up-to-date noisepropagation software to study the change in exposure due toincrement in existing traffic in a large panel of situations.Using such GIS software and data can allow the elaborationof spatialized and temporalized characterization factors. Thispossibility will be studied in future work.
7 Conclusions
This study aimed to compare the noise impact on humanhealth of two different tires, based on existing LCA method-ology. The noise characterization is still marginal in LCAstudies, and the present paper compares two very differentavailable methodologies, allowing the comparison of theirresults. Since noise impact category is still in development,it was important to understand the applicability and the out-comes given by the available methodologies. The large con-tribution of noise effects on overall human health results withFranco and Cucurachi methods highlights the necessity tointegrate noise impact on human health in LCA. More re-search and efforts should be encouraged, toward the study ofnoise impact, its integration in LCA, and its mitigation.
Franco and Cucurachi results showed an impact decreaseof 60.0 and 55.4 %, respectively, thanks to the technical char-acteristics of tire 2, as compared to tire 1. The outcomes areconsistent between the two methods because they are of thesame order of magnitude and they also show the improvementefforts made at the design phase to decrease the noise emis-sions. A good way to significantly reduce impact from noiseemissions seems to work on the design of the tires. The twomethods give a very different result in terms of DALY scoreswith more than eight orders of magnitudes between them. Theresult given by the Cucurachi method seems to beoverestimated with several DALYs by kilometers for one tire.
Both methods have advantages and drawbacks. However,it can be possible to choose a third method, in-between, com-bining the strength of the approach initiated by Müller-Wenkand the one imagined by Cucurachi. To do so, an incrementalapproach used on accurate localized and temporalized dataprocessed with noise propagation software could providecharacterization factors for a set of archetypes. This wayseems a good compromise between both approaches, and wewill work on it in the near future.
Int J Life Cycle Assess
Acknowledgments The authors gratefully acknowledge the GoodyearInnovation Center (GYIC) Luxembourg, in particular Dr. GeorgesThielen, for having co-funded the study and provided data for the twotires and Vanessa Peardon from LIST for English proofreading. Thisresearch has been funded by the Luxembourg National Research Fund(FNR) under the project DyPLCA (INTER/ANR/13/10/DyPLCA).
References
Althaus H-J, de Haan P, Scholz RW (2009) Traffic noise in LCA. Int JLife Cycle Assess 14:560–570
ChemRisk L (2009) Tyre generic exposure scenario; Service lifeguidance
Cucurachi S, Heijungs R (2014) Characterisation factors for life cycleimpact assessment of sound emissions. Sci Total Environ 468–469:280–291
Cucurachi S, Heijungs R, Ohlau K (2012) Towards a generalframework for including noise impacts in LCA. Int J LifeCycle Assess 17:471–487
De Hollander AE, Melse JM, Lebret E, Kramers PG (1999) An aggregatepublic health indicator to represent the impact of multiple environ-mental exposures. Epidemiol 10:606–617
Directive E (2002) Directive 2002/49/EC of the European parliament andthe Council of 25 June 2002 relating to the assessment and manage-ment of environmental noise. Off. J. Eur. Communities 189
Doka G (2003) Ergänzung der Gewichtungsmethode für ÖkobilanzenUmweltbelastungspunkte’ 97 zu Mobilitäts-UBP’ 97. DokaÖkobilanzen Zurich
Dutilleux G, Defrance J, Ecotière D, Gauvreau B, Bérengier M, BesnardF, Duc EL (2010) NMPB-routes-2008: the revision of the Frenchmethod for road traffic noise prediction. Acta Acust United Acust96:452–462
EASA (2009) Background noise level and noise level from en routeaircraft final report
ETRMA (2010) Replacement of highly aromatic oils in tyres – FAQ andDEFINITIONS [WWW Document]. URL http://www.etrma.org/uploads/Modules/Documentsmanager/20100712-etrma-qanda—replacement-of-ha-oils-in-tyres.pdf
Eurostat, European Commission [WWWDocument], n.d. URL http://ec.europa.eu/eurostat/ (accessed 2.26.16)
Evans L, MacIsaac Jr J, Harris J, Yates K, Dudek W, Holmes J, Popio J,Rice D, Salaani M (2009) NHTSA tire fuel efficiency consumerinformation program development: phase 2—effects of tire rollingresistance levels on traction, treadwear, and vehicle fuel economy.East Lib. OH Natl. Highw. Traffic Saf. Adm
Franco V, Garraín D, Vidal R (2010) Methodological proposals for im-proved assessments of the impact of traffic noise upon humanhealth. Int J Life Cycle Assess 15:869–882
French Road Safety reports, Observatoire National Interministériel de laSécurité Routière (ONISR) [WWW Document], n.d. URLhttp://www.securite-routiere.gouv.fr (accessed 2.26.16)
Frischknecht R, Jungbluth N, Althaus H-J, Doka G, Dones R, Heck T,Hellweg S, Hischier R, Nemecek T, Rebitzer G (2005) Theecoinvent database: overview and methodological framework. Int.J. Life Cycle Assess 10:3–9
Fritschi L, Brown L, Kim R, Schwela D, Kephalopolous S (2011)Conclusions [Burden of disease from environmental noise:Quantification of healthy years life lost in Europe]. World HealthOrganisation
GoedkoopM, Heijungs R, Huijbregts M, De Schryver A, Struijs J, van ZelmR (2009) ReCiPe 2008. Life Cycle Impact Assess. Method WhichComprises Harmon. Categ. Indic. Midpoint Endpoint Level 1
Heutschi K (2004) des forêts et du paysage Suisse. Office fédéral del’environnement, Eidgenössische Materialprüfungs-undForschungsanstalt. Akustik, Geinoz D, Liengme J-D, 2004.SonRoad: modèle de calcul du trafic routier. OFEFP
Hofstetter P, Müller-Wenk R (2005) Monetization of health damagesfrom road noise with implications for monetizing health impacts inlife cycle assessment. J Clean Prod 13:1235–1245
Huijbregts M, Hertwich EG, Reijnders L (2006) Including human healthdamages due to road traffic in life cycle assessment of dwellings. IntJ Life Cycle Assess 11:64–71
Institut national de la statistique et des études économiques (INSEE)[WWW Document], n.d. URL http://www.insee.fr/fr/ (accessed2.26.16)
ISO E (2006) 14044: 2006. Environ. Manag.-Life Cycle Assess.-Requir.Guidel. Eur. Comm. Stand
Krzyżanowski M, Kuna-Dibbert B, Schneider J (2005) Health effects oftransport-related air pollution. WHO Regional Office Europe
Lacour S, Joumard R (2002) Parc, usage et émissions des véhiculesroutiers en France de 1970 à 2020
Miedema HM, Oudshoorn CG (2001) Annoyance from transportationnoise: relationships with exposure metrics DNL andDENL and theirconfidence intervals. Environ Health Perspect 109:409
Müller-Wenk R (1999) Life-cycle impact assessment of road transportnoise. Institut für Wirtschaft und Ökologie, Universität St. Gallen(IWÖ-HSG)
Müller-Wenk R (2002) Attribution to road traffic of the impact of noiseon health environmental series no. 339: noise. Int J Life CycleAssess 7:376–376
Müller-Wenk R (2004) A method to include in lca road traffic noise andits health effects. Int J Life Cycle Assess 9:76–85
Nielsen P, Laursen J (2005) Integration of external noise nuisance fromroad and rail transportation in lifecycle assessment. Dan. Environ.Prot. Agency Cph
Noise Observation and Information Service for Europe (NOISE) [WWWDocument], n.d. URL http://noise.eionet.europa.eu/ (accessed2.26.16)
Ntziachristos L, Boulter P (2009) EMEP/EEA air pollutant emissionsinventory guidebook 2009: road vehicle tyre and brake wear; roadsurface wear. Copenhagen. Eur. Environ. Agency
Null V (1999) Safe process oils for tires with low environmental impact.Kgk Kautsch Gummi Kunstst 52:6
Observatoire de l’environnement sonore de la Métropole de Lyon[WWW Document], n.d.. acoucité. URL http://www.acoucite.org/(accessed 2.26.16)
PRé Consultants B (2001) Life cycle assessment of an average Europeancar tyre. Comm. Eur. Car Tyre Manuf. BLIC Third Party Rep. 23
Rauterberg-Wulff A (2003) Tire wear as source of PAH,Umweltbundesamt, Berlin CSTEE/2003/18/4
Steven H (2005) Ermittlung der Geräuschemission von Kfz imStraßenverkehr. TÜV Nord Würselen
Ten Broeke H, Hulskotte J, Van Der Gon H-D (2008) Emissionestimates for diffuse sources Netherlands EmissionInventory—road traffic tire wear. TNO Built Environmentand Geosciences. US department of energy (USDE) [WWWDocumen t ] . URL h t t p : / / www. em i s s i e r e g i s t r a t i e .nl/ERPUBLIEK/documenten/Water/Factsheets/English/Road%20traffic%20tyre%20wear.pdf (accessed 2.26.16)
a Luxembourg Institute of Science and Technology (LIST), 5, avenue des Hauts-FourneauxL-4362, Esch-sur-Alzette, Luxembourgb ETIS, UMR 8051, Université Paris Seine, Université Cergy-Pontoise, ENSEA, CNRS, 2 rue Adolphe Chauvin, 95302 Cergy-Pontoise Cedex, Francec Ifsttar, Laboratoire d’Acoustique Environnementale (LAE), Center of Nantes, CS4, 44344 Bouguenais Cedex, France
a r t i c l e i n f o
Article history:
Received 25 August 2016
Received in revised form 24 May 2017
Accepted 25 May 2017
Keywords:
Search radius
Noise prediction software
Environmental noise
Human health
a b s t r a c t
Engineering software products allow for quantifying environmental noise and a population’s exposure to
road traffic noise which can then be linked to human health damage. This paper investigates the impact
of the search radius, a parameter used in emission and propagation models, on noise exposure results.
The search radius is the threshold distance from which noise sources are not considered anymore in
the exposure assessment. To understand the influence of this parameter on the evaluation of population’s
exposure, the search radius has been successively fixed to three different values (500 m, 1000 m and
2000 m) in four different geographical situations (village, industrial, suburban and inner city). The result
of this investigation highlights several points. First, despite a search radius often fixed to 1000 m by noise
prediction software users, going up to 2000 m shows significant increase in population’s exposure.
Second, the impact of a change in search radius is very dependent of the presence of preponderant noise
sources. Third, increasing the search radius can quickly lead to an impractical calculation time. A solution
to avoid underestimating the exposure without increasing too much the calculation time may be to only
account for preponderant noise sources beyond a given distance.
� 2017 Elsevier Ltd. All rights reserved.
1. Introduction
Noise is defined as an unwanted sound. The impact of environ-mental noise on human health has been abundantly discussed byscientific communities in the last decades. The World HealthOrganisation (WHO) has provided a comprehensive review of thesestudies [1]. The impacts taken into account are cardiovascular dis-ease, cognitive impairment in children, sleep disturbance, tinnitusand annoyance. The WHO quantified the burden of disease fromenvironmental noise on human health, finding a range of burdenof 1.0–1.6 million disability-adjusted life year (DALYs) for WesternEurope. More than 90% of this amount is coming from sleep distur-bance and annoyance. This important burden of disease pushed theWHO to consider environmental noise as public health problem.
There are several sources for environmental noise, such astransportation (road, railway and aircraft traffic), industrial activi-ties, construction work, energy resources (wind turbine), andleisure activities. Among them, transportation, especially road traf-fic, is predominant. For example, in France, it has been estimated
that the health costs due to environmental noise are largely causedby road traffic (89%) [2]. For this reason, this paper focuses only onroad traffic.
In order to assess the impact of noise on human health, it is nec-essary to evaluate the population’s exposure. The European direc-tive 2002/49/CE [3] requires the assessment and management ofthe environmental noise for major European cities. One of the fast-est and most efficient ways to evaluate exposure is to generatenoise maps with prediction software using simplified sound prop-agation models. There are various noise emission and propagationmodels that can be used, including the ones developed under theEuropean projects IMAGINE or HARMONOISE and the FrenchNMPB08 method [4–7]. NMPB08 has been preferred because it ismore recent and more accurate than previous noise emission andpropagation models, as described in Ecotiere et al. [8]. Engineeringsoftware products allow for quantifying exposure to environmen-tal noise coming from road traffic for a given population (as wellas noise coming from railway traffic and industrial activities). Theyalso allow to take into account different scenarios such as the pres-ence of sound barriers, absorbing grounds, changes in road surfacecharacteristics, speed limits, and changes in traffic.
Once the exposure of the population has been calculated, itcan be linked with impacts on human health by following the
http://dx.doi.org/10.1016/j.apacoust.2017.05.028
0003-682X/� 2017 Elsevier Ltd. All rights reserved.
⇑ Corresponding author at: Luxembourg Institute of Science and Technology
(LIST), 5, avenue des Hauts-FourneauxL-4362, Esch-sur-Alzette, Luxembourg.
recommendations of the WHO [1]. In this paper, only the two majorimpacts will be calculated, i.e. annoyance and sleep disturbance. Thelink between exposure and human health is the percentage of highlyannoyed (HA) persons, in the case of annoyance, and the percentageof highly sleep disturbed (HSD) persons, in the case of sleepdisturbance. Knowing the exposure, the number of persons highlyannoyed and/or highly sleep disturbed can be calculated using thedose response curves given, respectively, by Miedema andOudshoorn [9] and Miedema and Vos [10].
When implementing this approach, several parameters have tobe optimized, e.g. the search radius. The search radius defines a cir-cle around a receiver point, where the sound sources inside thiscircle will be considered in the calculation while the sound sourcesoutside this circle will be neglected. It can be seen as the allowed‘‘maximum propagation distance”. Studying the impact of thissearch radius on population exposure assessment is the mainobjective of this paper.
2. Material and methods
2.1. Population exposure assessment
In this study, a noise prediction software called CadnaA [11] isused. Among the calculation parameters of CadnaA, the so called‘‘Max. Search Radius” is the parameter of interest for this paper.This parameter is also called ‘‘maximum path length” in the NMPB2008 methodological guide [12]. For the sake of simplicity, thissearch radius will be referenced as dmax. When taking the pointof view of a building, all roads within the dmax will be consideredto evaluate the noise level at which this building is exposed. Thevalue chosen for dmax can have a large impact on the noise levelcalculated on each building’s facade.
Despite the sensitivity to dmax, there is no imposed or even rec-ommended value for this parameter in the European directive2002/49/EC [3]. In most cases, a typical value of 1000 m is used[13,14]. According to the NMPB 2008, the method used in thispaper for the configuration of CadnaA, the dmax is valid up to2000 m [12].
On the one hand, a higher value for the maximum propagationdistance implies that more sources will be considered in the calcu-lation of the noise level on the studied façades. If the noise propa-gation software works properly and the simulation is well done, aresult with a higher dmax should give a result more representativeof the reality. On the other hand, the number of considered sourceswill grow at the same rate as the square of dmax since the numberof considered sources is proportional to the surface taken intoaccount in the calculation (assuming a homogeneous distributionof sources). Moreover, the dmax parameter also modifies the num-ber of potential reflection and ray paths. Thus the calculation timecan quickly become unpractical.
Since a value of 1000 m is the typical value used by peopleworking on noise maps [13,14], a dmax value of 1000 m is chosenas a reference point. It may be interesting to compare two differentchanges in the value of the search radius to evaluate the benefits offixing dmax at 2000 m instead of the mostly used value of 1000 m.As a result, it has been chosen to compare a doubling of dmax from500 m to 1000 m and from 1000 m to 2000 m.
In order to study the impact of the search radius, GeographicInformation System (GIS) data was used. It is given by a localFrench agency, Acoucité [14], and contains all the necessary infor-mation for the exposure assessment (roads, traffic, buildings,inhabitants, topography, ground characteristics, etc.). The area ofthe study is the Grand Lyon region that corresponds to the Metro-polis of Lyon, a French territorial authority. The GIS data has beenmanipulated with a free and open-source software: OrbisGIS [15].
For the spatial scale of the study, a geographical mapping of theFrench territory called IRIS (Aggregated Units for Statistical Infor-mation - Ilots Regroupés pour l’Information Statistique) was chosen.
IRIS is the basic unit for the collection and transmission of sta-tistical data coming from the French National Institute of Statisticsand Economic Studies (INSEE) [16]. These geographical areas(16100 IRIS in France in total) are on a district scale and containbetween 1800 and 5000 inhabitants. More importantly, IRIS arebuilt in a way to ensure homogeneity among geographic anddemographic criteria [17], and that is the main reason why theyhave been chosen for this analysis. Moreover, while IRIS are largeenough to contain hundreds of buildings to be evaluated for eachsituation, they are small enough to allow for a lot of different cal-culations. This leads to a large amount of results which could bestatistically analysed.
For each studied IRIS, all of the inhabited buildings in the areacontained in the IRIS itself and an additional buffer of 1000 m havebeen evaluated for three values of search radius (dmax = {500 m;1000 m; 2000 m}). The buffer was chosen to have a higher numberof evaluated buildings to ensure a real ‘‘signal” and not just some‘‘noise”. The purpose of the buffer was to add more points in thestudied cases by considering a larger area.
For each evaluated building, the noise level, L, is the maximumnoise level found at four meters above ground and at two meters infront of all the façades of the studied building, following the stan-dard of European directive 2002/49/CE [3]. The noise predictionsoftware gives noise levels Lday, Levening, Lnight and Lden. Lday, Leveningand Lnight that correspond, respectively, to the noise level duringday (6–18 h), evening (18–22 h), and night (22–6 h), while Lden isa day-evening-night equivalent level.
The few existing typologies for IRIS seem to be based on socio-economic factors (e.g. [18]). To our knowledge, there is no existingtypology based on environmental noise at the IRIS scale. There aresome typologies based on noise for urban situations, but they areat the street scale [19,20]. Nevertheless, the aim of this paper isnot to establish a noise typology of IRIS, but to show the influenceof the search radius on different types of IRIS. The studied IRIS havebeen chosen to cover the different possibilities with a relativelylow sampling rate. Four IRIS have been selected which are consid-ered to be representative of the different types of configurationsencountered in the Grand Lyon area. This choice regarding the geo-graphical areas will be discussed in Section 4.1. For the sake of sim-plicity, a nickname is given for each of the four studiedgeographical area: Village, Industrial, Residential and City. It has tobe noted that, as the four IRIS do not represent prototypes of anytypology, the so-called Village is a geographical area which is notnecessarily representative of every village.
In Fig. 1 representing the Village, the space between the twooutlines is the 1000 m buffer. It means that the noise level in frontof all the inhabited buildings contained inside the exterior outlinewill be predicted. All the black squares are buildings while the darkcurving lines are roads. Finally, the grey lines are contour lines rep-resenting the topography. The noise prediction model will searchfor all the noise sources at less than dmax from the studied building.
The Village is centred on a small town in a hilly landscape.Within the buffer and the search radius, the considered area ismore heavily urbanised than the IRIS alone, but it is one of the lessdensely inhabited areas in the Grand Lyon region. There is a rail-way line, but the road traffic is the only noise source consideredin this work. There are no major roads in the area, so the environ-mental noise is expected to be low.
In the Industrial area, Fig. 2, most of the buildings in the IRIS areuninhabited. This is because it is an industrial area with big build-ings and open space. Once again, railways are present but nottaken into account. It means that most of the inhabited buildingsthat will be evaluated are not in the IRIS itself but in the buffer.
64 R. Meyer et al. / Applied Acoustics 127 (2017) 63–73
The landscape is relatively flat. Some roads present very dense traf-fic, so the environmental noise is expected to be high.
In Fig. 3, most of the buildings in the IRIS and its buffer are res-idential houses. This small town is surrounded by crops. There is a
road with heavy traffic crossing the IRIS; in this flat landscape, thesound will propagate without encountering any obstacles, so theenvironmental noise is expected to be high. The south of the eval-uated zone is mostly uninhabited buildings.
Fig. 1. Geographical area nicknamed ‘‘Village”. The interior bold outline is the IRIS, and the exterior bold outline delimits the 1000 m buffer. The black squares are buildings,
the black lines define the road network, and the grey ones represent the topography.
Fig. 2. Geographical area nicknamed ‘‘Industrial”. The interior bold outline is the IRIS, and the exterior bold outline delimits the 1000 m buffer. The black squares are
buildings, the black lines define the road network, and the grey ones represent the topography.
R. Meyer et al. / Applied Acoustics 127 (2017) 63–73 65
Fig. 3. Geographical area nicknamed ‘‘Residential”. The interior bold outline is the IRIS, the exterior bold outline delimits the 1000 m buffer. The black squares are buildings,
the black lines define the road network, and the grey ones represent the topography.
Fig. 4. Geographical area nicknamed ‘‘City”. The interior bold outline is the IRIS, the exterior bold outline delimits the 1000 m buffer. The black squares are buildings, the
black lines define the road network, and the grey ones represent the topography.
66 R. Meyer et al. / Applied Acoustics 127 (2017) 63–73
In Fig. 4, one can see a very densely urbanised area. This isdowntown Lyon, hence the high population density. This is a flatterrain with a lot of roads, but with the speed limit in city, onecan expect medium environmental noise for such an area.
The most important characteristics of the four studied IRIS aresummarised in Table 1.
2.2. Environmental noise impact on human health
Using the CadnaA prediction software (Section 2.1), the expo-sure of the population to environmental noise is obtained throughLday, Levening, Lnight and Lden index. The WHO [1] quantifies the bur-den of disease using the DALY indicator defined by the followingequation:
DALY ¼ YLLþ YLD ð1Þ
This indicator combines two different metrics, the years of lifelost due to premature death (YLL) and the years lived with disabil-ity (YLD). The YLL are expressed by:
YLL ¼ N � L ð2Þ
where N is the number of deaths due to the studied condition, and Lis the standard life expectancy at age of death (expectancy – age atdeath). In this paper, the two major contributors to the DALYs com-ing from environmental noise are annoyance and sleep disturbance.In these two cases, there will not be any premature death, soYLL = 0. YLD can be estimated with the following equation:
YLD ¼ I � DW � D ð3Þ
where I is the number of incident cases for the studied condition,the disability weight (DW) reflects the severity of the disability ona scale from zero (no adverse health effects) to one (equivalent todeath), and D is the average duration of the disability in years.
Accounting only for annoyance and sleep disturbance, thehuman health impact calculated in this paper with the followingequation:
HH ¼ NHA � DWHA þ NHSD � DWHSD ð4Þ
NHA is the number of highly annoyed persons and NHSD the numberof highly sleep disturbed persons. The disability weights for annoy-ance (DWHA) and sleep disturbance (DWHSD) can be found in thestudy of the WHO [1] and are reported in Table 2.
The only step missing between population exposure to environ-mental noise and human health impact is calculating the numberof HA and HSD persons. The relationship from Miedema and Oud-shoorn [9] can be applied to obtain the number of highly annoyedpersons knowing noise exposure in terms of Lden for each buildingand the number of inhabitants. It is given by:
The relationship from Miedema and Vos [10] can be used toobtain the number of highly sleep disturbed persons knowing theexposure in terms of Lnight for each building and the number ofinhabitants:
Using these formulas, the human health impact of environmen-tal noise can be linked to population exposure and calculated foreach of the four cases exposed in Section 2.1.
3. Results
3.1. Qualitative approach for noise levels
The change in Lden values resulting from a doubling of dmax iscalculated. In order to visualize this change for each evaluatedbuilding, the increase in the evaluated noise level is plotted versusthe noise level evaluated for the smaller value of dmax. The resultsare given for each geographical studied area in Fig. 5.
Fig. 5(a) and (b) show important increases in noise level for aninitial noise level between 25 and 40 dB(A). In Fig. 5(c–f), the mostimportant increases in the noise level are occurring between 40and 60 dB(A). As expected, environmental noise is lower in thesmall village than in IRIS near major roads.
Fig. 5 also shows another anticipated trend: the maximum dif-ference between the noise levels obtained when doubling dmax
decreases with the initial noise level. In other words, lower noiselevels are more sensitive to a change in maximum propagation dis-tance than higher noise levels. A given amount of acoustical powerwill lead to a higher increase in noise level when starting from alower initial noise level. This is a direct consequence of the loga-rithmic scale on which the decibel is built. The increase in noiselevel can be very important for lower initial noise level, as seenin Fig. 5(c), (e) and (f), where increases can exceed 10 dB(A).
It can be also observed that the four selected geographical areasrepresent different behaviours. Fig. 5(a–f) show a higher increasein noise level between 500 m and 1000 m than between 1000 mand 2000 m. On the contrary, (g) and (h) show a significantincrease in the noise levels for some of the evaluated buildingswhen the maximum propagation distance passes from 1000 m to2000 m but none for a change from 500 m to 1000 m. This beha-viour will be discussed later in Section 3.4.
3.2. Quantitative approach for noise levels
With the first approach considered in Fig. 5, it is hardly possibleto see the percentage of buildings undergoing an increase in noiselevel, as thousands of points are plotted. Moreover, an increase canonly be considered as relevant if it occurs at a noise level that hasan impact on human health and if it exceeds a certain threshold. Inthis study, a threshold of 1 dB(A) has been chosen, considering thatan increase of less than 1 dB(A) is negligible as it cannot be
Table 1
Main characteristics of selected IRIS.
Nickname IRIS Description Number of evaluated
buildings
Studied population Density
(inh/km2)
Village 692330000 Small village in a valley. 2408 9832 800
Industrial 692730101 Industrial zone and suburban area, near a road with dense traffic. 2865 14737 820
Suburban 692710104 Crops and residential area crossed by a road with dense traffic. 3563 29223 920
City 693830501 Heavily urbanised, in downtown Lyon. 4043 93923 18 000
Table 2
Disability weight for annoyance and sleep disturbance from the WHO.
Health problem Disability weight (DW)
Annoyance 0.02
Sleep disturbance 0.07
R. Meyer et al. / Applied Acoustics 127 (2017) 63–73 67
detected by the human ear, thus it is not worth the computationaleffort and calculation time.
Identifying the minimum sound level to be considered is diffi-cult. Based on the WHO [1], the lower limit is 45 dB(A) for Lnight
and 55 dB(A) for the Lden. This last part involves knowing exposureabove 45 dB(A) for Lnight, above 50 dB(A) for Levening, and above55 dB(A) for Lday. In order to show the impact of a change in dmax
related to the noise level exposure in a quantitative way, the
Fig. 5. Noise level increase for a doubling of dmax for our four IRIS (a) village, going from dmax = 500 m to dmax = 1000 m, (b) village, going from dmax = 1000 m to
dmax = 2000 m, (c) industrial, going from dmax = 500 m to dmax = 1000 m, (d) industrial, going from dmax = 1000 m to dmax = 2000 m, (e) suburban, going from dmax = 500 m to
dmax = 1000 m, (f) suburban, going from dmax = 1000 m to dmax = 2000 m, (g) city, going from dmax = 500 m to dmax = 1000 m, (h) city, going from dmax = 1000 m to
dmax = 2000 m.
68 R. Meyer et al. / Applied Acoustics 127 (2017) 63–73
relative number of evaluated buildings above these thresholdsundergoing a change in exposure higher than 1 dB(A) has been cal-culated. The results are given in Table 3.
With this second approach, it is possible to detect two differentbehaviours. In Village and City areas, the increase in noise levelexposure due to changes of dmax is very low. In these two IRIS, lessthan 5% of the buildings exposed to noise levels above 45 dB(A)undergo an increase superior to 1 dB(A) and none for the oneexposed to noise levels above 55 dB(A). However, in Industrial
and Suburban areas, a significant part of the evaluated buildingundergoes an increase of more than 1 dB(A), especially between45 dB(A) and 55 dB(A). In these two IRIS, a doubling of dmax hasmore impact on Lnight than on Lday or Levening. The night periodmay be more sensitive to a change in search radius because the dif-ference between heavily trafficked roads and the environmentalnoise due to the rest of the road network may be more important.
3.3. Impact on human health
In order to quantify the impact of environmental noise, it can beinteresting to push the analysis one step further. Following the rec-ommendation of the WHO [1]: the number of HA persons and thenumber of HSD persons can be calculated in these four geograph-ical areas (see Eqs. (5) and (6)). Knowing the numbers of HA andHSD persons in each case allows the calculation of the DALYs
due to environmental noise as detailed in Section 2.2. These threeindicators have been calculated for each maximum propagationdistance (Table 4).
The human health impact in Table 4 is calculated using Eq. (4)and the disability weights given in Table 2. Studying the impor-tance of HA and HSD on the human health impact in DALYs, theshare of HSD in the DALYs can be calculated as explained in the fol-lowing equation:
Share of HSD in the DALYs ¼NHSD � DWHSD
NHA � DWHA þ NHSD � DWHSD
ð7Þ
The bigger contribution comes from HSD ranging between 58%(Industrial, dmax = 2000 m) and 67% (Suburban, dmax = 500 m).While averaging on the whole table (12 values), the mean shareof HSD in the DALYs is 61%. Even if there are more HA persons,the higher disability weight for HSD is compensating. The studiedpopulation changes considerably between the Village and the City,which partially explains why there is a difference in the DALYs’amount. Dividing the amount in DALYs at dmax = 2000 m by thepopulation gives values ranging from 0.0022 DALYs/person forthe Village to 0.0087 DALYs/person for the City. On average, some-one living in the studied City loses 3.2 days of ‘healthy” life per yearwhile one living in the Village loses 0.79 days of ‘‘healthy” life peryear. It is not surprising to find a higher human health impact in
Table 3
Percentage of evaluated buildings in the four IRIS undergoing an increase in noise level exposure higher than 1 dB.
Number of highly annoyed persons, highly sleep disturbed persons and DALYs for each geographical area studied and each maximum propagation distance.
Geographical area Village Industrial Suburban City
Population 9832 14737 29224 93924
Number of highly annoyed persons dmax = 500 m 411 727 1655 15601
dmax = 1000 m 414 858 2200 15615
dmax = 2000 m 416 1010 2679 15641
Number of highly sleep disturbed persons dmax = 500 m 183 365 975 7207
dmax = 1000 m 183 380 1081 7213
dmax = 2000 m 183 398 1150 7227
Human Health Impact (DALYs) dmax = 500 m 21 40 101 816
dmax = 1000 m 21 44 120 817
dmax = 2000 m 21 48 134 819
R. Meyer et al. / Applied Acoustics 127 (2017) 63–73 69
a louder situation. However, the amount of days of healthy life lostis important and explains the concerns related to environmentalnoise.
Regarding the WHO report [1] and taking only into accountannoyance and sleep disturbance, between 922,000 and1,490,000 DALYs have been found for a population of 285 million.The numbers from this WHO report lead to an impact per personranging from 0.0032 DALYs/person to 0.0052 DALYs/person. Thequietest area studied in this work (0.0022 DALYs/person) is underthe mean for Europe while the noisy city (0.0087 DALYs/person) isover that mean value. Given the limited sampling of this study, itmakes no sense to evaluate an average, but the results havenonetheless the same order of magnitude.
One way to understand the importance of the amount found inDALYs concerning environmental noise is to take a look at theimpact of air pollution. Van Zelm [21] considered the humanhealth effects of fine particulate (PM10) and ozone in Europe. Theyfound 0.003 DALYs/person for a European average. It means thatthe impact on human health coming from environmental noiseis, to the current knowledge of scientists, in the same order of mag-nitude as the impact of fine particulate and ozone.
The focus of this work is the assessment of the increase of thethree indicators (HA, HSD, DALY), while increasing the maximumpropagation distance. In order to better understand the importanceof this increase, the results are normalized and exposed in Table 5.For example, when looking at the increase in the number of HApersons when dmax goes from 500 m to 1000 m, the percentageof HA people at 1000 m explained by the change will be (HA(1000 m)-HA(500 m))/HA(1000 m).
Table 5 shows two main results. First, in the Industrial andSuburban areas, the variation of the three indicators when increas-ing the maximum propagation distance is important. Up to 10% ofthe impact in DALYs can be explained by the change in dmax fromdmax = 1000 m to dmax = 2000 m. Thus the impact of increasing themaximum propagation distance to 2000 m can be very importantwhen evaluating environmental noise impact on human health.One has to keep this information in mind while looking at a DALYscore. It seems sensitive to parameters that are not always dis-cussed or even stated when the results of population exposureassessment are given (e.g. in the case of the application of the Euro-pean directive 2002/49/CE [3]).
The other major result is the presence of two different beha-viours. Table 5 shows normalized increases between 4% and 24%for the three indicators in both Industrial and Suburban areas,whereas there is almost no increase for Village and City areas. Theseresults are in agreement with the observations made previously(see Fig. 5 and Table 3). Heavily trafficked roads can explain thelarge difference between these two behaviours, as developed inthe next section.
3.4. Heavily trafficked roads
Table 1 shows that the geographical areas most affected by theincrease in the maximum propagation distance are the two that
contain a heavily trafficked road. Indeed, the Industrial and Subur-
ban area contain, or are adjacent to, roads with a daily trafficgreater than 2000 passenger cars per hour at a speed of 90 km/h.Fig. 6 presents all the roads with an hourly traffic higher than2000 passenger cars. The interior, bold, black outline is the IRIS,while the exterior, bold, black outline delimits the 1000 m buffer.The red, orange and yellow areas are areas located at less than500 m, less than 1000 m, and less than 2000 m, respectively, froma heavily trafficked road.
These heavily trafficked roads are the most important noisesources compared to the other roads taken into account in theexposure calculation. That is why they have an important impacton the noise level evaluated at buildings’ facades. If the changein the maximum propagation distance adds a heavily traffickedroad that was not considered before (a building in an orange or yel-low zone on Fig. 6), the noise exposure at the building undergoesan important increase (see Fig. 5(c–f) and Table 3 concerningIndustrial and Suburban). Fig. 6(b) and (c) show this phenomenonvery well with a large number of buildings in the orange and yel-low areas. If all the studied buildings are located under 500 m (inthe red area) or over 2000 m (in the white area) of all heavily traf-ficked roads, changing the search radius would not lead any build-ings to consider a heavy traffic road that was not consideredbefore, so it would change almost nothing in the exposure results.
Concerning the City, Fig. 5(g) and (h) and Table 3 show a similarbut less significant effect. Fig. 6(d) shows that there are less build-ings in the orange zone than in the yellow zone. That explains whya cluster of points can be seen well above the baseline (at 0 dB(A)),when dmax goes from 1000 m to 2000 m but not when dmax goesfrom 500 m to 1000 m. However, when looking at Fig. 6(d), onecan be surprised that the large amount of buildings in the orangezone do not have a higher impact compared to Industrial andSuburban. This difference is explained here by the fact that heavilytrafficked roads in this area are limited to 50 km/h and the sur-rounding noise is higher. Moreover, the high density of buildingsin the city creates a ‘‘mask effect”, which blocks some of the noisethat could reach another building’s façades.
If an IRIS is in the neighbourhood of a heavily trafficked road,the most influential parameter when assessing the importance ofa change of the search radius is the distance between studiedbuildings and heavily trafficked roads. The results for Village inTable 3 show that in the absence of heavily trafficked roads(Fig. 6(a)), the impact of a change in the search radius seems tobe lower when dmax goes from 1000 m to 2000 m than when dmax
goes from 500 m to 1000 m.
3.5. Calculation time
The drawback of a higher dmax is its longer calculation time, aresult of taking more roads into account. The calculation timesfor the four IRIS considered and different values of dmax are givenin Table 6.
Table 6 shows that the calculation time for a dmax value of2000 m is much higher than dmax = 1000 m, roughly a day instead
Table 5
Percentage of HA persons, HSD persons, and DALYs explained by the change in the search radius.
IRIS Village Industrial Suburban City
Increase in HA persons From 500 to 1000 m 0.8% 15.3% 24.8% 0.1%
From 1000 to 2000 m 0.3% 15.0% 17.9% 0.2%
Increase in HSD persons From 500 to 1000 m 0.4% 4.1% 10.8% 0.1%
From 1000 to 2000 m 0.0% 4.8% 6.4% 0.2%
Increase in DALYs From 500 to 1000 m 0.6% 8.4% 15.3% 0.1%
From 1000 to 2000 m 0.1% 9.0% 10.8% 0.2%
70 R. Meyer et al. / Applied Acoustics 127 (2017) 63–73
of few. If one needs to perform a lot of calculations or to calculate apopulation’s exposure over a large geographical area, choosingdmax = 2000 m may be impractical.
The calculation time per building varies with different geo-graphical areas. On the one hand, the Village area is a complex
landscape with a complex topography which implies a longer cal-culation of the noise propagation. In the other hand, the Suburban
area is significantly lower because it is a relatively flat landscapeand not densely urbanised.
In theory, due to the 2D geometry of the problem, the calcula-tion time per building should grow with the square of dmax if theroad network is uniform, growing in the same pattern as the sur-
face considered (p � d2maxÞ. If one plots the calculation time per
buildings, one will see that it follows roughly what was theoreti-cally predictable.
4. Discussion
4.1. General consideration
This study was carried out in a limited number of areas. Even ifthey were selected to represent very different situations, theyprobably do not represent all the possibilities. The Village,Industrial, Suburban and City areas likely do not represent everyvillage, industrial area, suburban residential area, or city found inFrance or around the world. However, to study a limited amount
Fig. 6. Influence of heavily trafficked roads. IRIS and buffer (bold outlines). Areas located at less than 500 m, 1000 m and 2000 m from heavily trafficked roads (respectively
red, orange and yellow areas). (a) Village, (b) industrial, (c) suburban, (d) city. (For interpretation of the references to colour in this figure legend, the reader is referred to the
web version of this article.)
Table 6
Calculation time for different values of dmax per the four IRIS studied.
Geographical
area
dmax
(m)
Calculation time
(min)
Calculation time by
buildings (s)
Village 500 85 2.1
1000 508 12.7
2000 1479 36.9
Industrial 500 68 1.4
1000 302 6.3
2000 959 20.1
Suburban 500 50 0.8
1000 184 3.1
2000 826 13.9
City 500 63 0.9
1000 241 3.6
2000 1548 23.0
R. Meyer et al. / Applied Acoustics 127 (2017) 63–73 71
of geographical area is not a problem because it is sufficient in find-ing some specific behaviours. A more exhaustive approach wouldbe necessary to validate the observations.
This study shows that the higher noise levels are more resistanttowards a change in the search radius than lower noise levels.Looking at the impact on human health coming from environmen-tal noise for low levels, such as 45 dB(A) or even 50 dB(A), can betricky because of the sensitivity of these noise levels to simulationparameters. The uncertainties regarding lower noise levels will besubstantially higher than the ones of high noise levels because oftheir higher sensitivity. These important changes in noise levelswhen changing the search radius are also visible in the results inDALYs, i.e. on the human health impacts of environmental noiseas it is calculated today. A practitioner calculating the burden ofdisease attributable to environmental noise must keep this in mindand interpret the results accordingly.
In the presence of heavily trafficked roads in the area of study orin the surroundings, the chosen search radius will have a stronginfluence on the result. The presence of a noise source strongerthan the majority of the noise sources in the studied zone can sig-nificantly affect the exposure of the evaluated buildings, particu-larly when a change in the search radius integrates a heavilytrafficked road in the calculation. To account for this, the searchradius should be adjusted according to different situations. If thearea has a heavily trafficked road, the search radius should beincreased to the possible maximum while it can be decreased inareas where the emission sound power level of the noise sourcesis relatively homogeneous. Applying this kind of rule on the GrandLyon area would imply the use of a search radius of 2000 m innearly all the IRIS because there are only few cases where thereis absolutely no influence from heavily trafficked roads. However,using 2000 m instead of 1000 m will considerably increase thetime of calculation (see Section 3.5).
A more elegant solution for this particular problem could be toassociate a search radius to each road, depending on the emissionnoise level of this specific road in absolute terms or compared tothe other noise sources in the surroundings. Above a given dis-tance, an evaluated building will not consider all the noise sourcesbut only the strongest, i.e. the heavily trafficked roads. Thismethodology would have the advantage of being more economicin terms of calculation effort, but it would need a change in theapproach used in the environmental noise prediction software. Inthe selected software, CadnaA, the same search radius is used foreach source without consideration for the relative importance ofa noise source compared with its surroundings.
Beyond this paper, the goal of this research is to integrate theimpact of environmental noise on human health in the frameworkof life cycle assessment (LCA). The need for accuracy and represen-tativeness is not the same in LCA as it is in acoustics: only a simpleand robust method is needed. A non-systematic method could becriticized for the possible biases coming from the operator’schoices. Nothing in this study justifies choosing a search radiusof 1000 m. On the contrary, the results presented here push towarda search radius of 2000 m. There is no theoretical reason to use admax of 1000 m instead of 2000 m, but there is a practical reason.A lot of calculations cannot be feasible with a search radius of2000 m.
4.2. Proposition
Taking a search radius of 2000 m can lead to high calculationtime. Since the presence of heavily trafficked roads seems to bethe most important parameter, these roads can explain a large partof the difference in the noise levels of exposure when the maxi-mum propagation distance is increased. Therefore, a good compro-mise could be to take into account the road networks at less than1000 m from the studied buildings and then only the major roadsbetween 1000 m and 2000 m.
In order to identify the feasibility of such proposition, an IRIShas been chosen where most buildings are located at a distancebetween 1000 m and 2000 m from all major roads. This IRIS is dif-ferent from the other ones in this paper and has been chosen onlyto satisfy this condition. All the 641 buildings in this IRIS are eval-uated. The whole population contained in this geographical area is1765 persons. The buildings and landscape have been picked insidea buffer of 2000 m around this area. The road network has beenentirely selected up to 1000 m and, depending on the case, thewhole road network or only the major roads between 1000 mand 2000 m have been picked. Three simulations were run: (1) amaximum propagation distance of 1000 m, (2) a maximum propa-gation distance of 2000 m and the whole road network, and (3) amaximum propagation distance of 2000 m but only major roadsbetween 1000 and 2000 m. The resulting calculation time is givenin Table 7 along with the three indicators studied in the previoussection: number of HA persons, number of HSD persons, andDALYs.
As expected, the three indicators and the calculation timeincrease with dmax. Taking into account all the roads instead of onlyheavily trafficked ones lead also to higher results, both in humanhealth impacts and calculation time. The calculation time ofScenario 2 is 4.6 times higher than Scenario 1. Scenario 3 providesa compromise between the two, with a calculation times 2.7 timeshigher than Scenario 1, but also greater coverage of noise-relatedhuman health impacts. Applying Scenario 3 instead of 2 leads saves40% of the calculation time. Additionally, Scenario 3 explains 76%of the change between Scenarios 1 and 2 in the number of HA per-sons, 82% of the change in the number of HSD persons, and 77% ofthe change in DALYs. In other words, considering only major roadsbetween 1000 m and 2000 m adds only half of the supplementarycalculation time but explains more than three quarters of thedifference in terms of HA persons, HSD persons or DALYs.
This small experiment is only qualitative because of technicalreasons. The buildings in the middle of the IRIS take into accountthe whole road network up to 1000 m plus their distance to theborder of the IRIS in each case. So this experiment is only a roughapproach to test this hypothesis on the importance of heavily traf-ficked roads with the available tools. Taking only one building intoaccount leads to a too quick calculation time to allow comparisonbetween scenarios, so this calculation had to be carried out on alarger scale. However, even doing the experiment on the single IRISand in an approximate manner, it shows a good compromise toobtain a more precise noise level evaluation without increasingthe calculation time too drastically. This proposed methodologymay be a path worth exploring to refine the calculations of noiseprediction software.
Table 7
Calculation time in the three different scenarios.
Scenario dmax (m) Road considered between 1000 m and 2000 m from the IRIS Calculation time (s) HA persons HSD persons DALYs
1 1000 None 1231 53.5 32.5 3.34
2 2000 All 5624 78.7 34.7 4.00
3 2000 Only roads with more than 2000 passenger cars by hour 3358 72.7 34.3 3.85
72 R. Meyer et al. / Applied Acoustics 127 (2017) 63–73
5. Conclusion
Studying the impact coming from a change in search radiushighlights several interesting elements. A search radius is oftenfixed to 1000 m by noise prediction software users, but a signifi-cant increase in the noise level evaluated can occur in some situa-tions when setting the search radius at 2000 m.
The impact of this change in the maximum propagation dis-tance is very dependent on the presence of one or several prepon-derant noise sources, such as a heavily trafficked road in a calm/quiet area. However, setting the search radius at 2000 m can beproblematic for practical reasons. In particular, the additional costin terms of calculation time may not be sustainable. Choosing themaximum distance propagation on a case to case basis can alsolead to criticism over the choices made by the operator.
An accommodating way to solve this problem could be to havea search radius depending on the emission noise level of the noisesources in absolute or relative terms. Noise prediction softwaremay look in this direction to integrate heavily trafficked roads inthe evaluation of noise exposure, allowing a more accurate evalu-ation of human health impacts without increasing too much thecalculation time.
Acknowledgement
This research has been funded by the Luxembourg NationalResearch Fund (FNR) under the project DyPLCA (INTER/ANR/13/10/DyPLCA).
References
[1] Fritschi L, Brown L, Kim R, Schwela D, Kephalopolous S. Conclusions [Burden ofdisease from environmental noise: Quantification of healthy years life lost inEurope]. World Health Organisation; 2011.
[3] Directive E. Directive 2002/49/EC of the European parliament and the Councilof 25 June 2002 relating to the assessment and management of environmentalnoise. Off J Eur Commun 2002; 189(12).
[4] Dutilleux G et al. NMPB-Routes-2008: the revision of the french method forroad traffic noise prediction. Acta Acust. United Acust 2010;96(3):452–62.
[5] IMAGINE. Guidelines for the use of traffic models for noise mapping and noiseaction planning. vol. Deliverable 7 of the IMAGINE (Improved methods for theassessment of the generic impact of noise in the environment) Project, TNOTechnical Report, Delft, The Netherlands; 2007.
[6] IMAGINE. The noise emission model for European road traffic. vol. Deliverable11 of the IMAGINE (Improved methods for the assessment of the genericimpact of noise in the environment) Project, M+P—Consulting EngineersTechnical Report, Vught, The Netherlands; 2007.
[7] van Maercke D, Defrance J. Development of an analytical model for outdoorsound propagation within the harmonoise project. Acta Acust United Acust2007;93(2):201–12.
[8] Ecotiere D, Foy C, Dutilleux G. Comparison of engineering models of outdoorsound propagation: NMPB2008 and Harmonoise-Imagine. Acoustics 2012;2012.
[9] Miedema HM, Oudshoorn CG. Annoyance from transportation noise:relationships with exposure metrics DNL and DENL and their confidenceintervals. Environ Health Perspect Apr. 2001;109(4):409.
[10] Miedema HME, Vos H. Associations between self-reported sleep disturbanceand environmental noise based on reanalyses of pooled data from 24 studies.Behav Sleep Med 2007;5(1):1–20.
[13] Bruitparif, le bruit en Ile-de-France. <http://www.bruitparif.fr/> [accessed: 04-Jul-2016].
[14] Observatoire de l’environnement sonore de la Métropole de Lyon. acoucité.<http://www.acoucite.org/> [accessed: 26-Feb-2016].
[15] OrbisGIS official website. <http://orbisgis.org/> [accessed: 04-Jul-2016].[16] Institut national de la statistique et des études économiques (INSEE). <http://
c1523> [accessed: 30-Mar-2017].[18] Asterop : Le géomarketing décisionnel. <http://www.asterop.fr/fr/
produits/donnees/typologies_IRIS.html> [accessed: 04-Jul-2016].[19] Wang B, Kang J. Effects of urban morphology on the traffic noise distribution
through noise mapping: a comparative study between UK and China. ApplAcoust 2011;72(8):556–68.
[20] Silva LT, Oliveira M, Silva JF. Urban form indicators as proxy on the noiseexposure of buildings. Appl Acoust 2014;76:366–76.
[21] van Zelm R et al. European characterization factors for human health damageof PM10 and ozone in life cycle impact assessment. Atmos Environ 2008;42(3):441–53.
R. Meyer et al. / Applied Acoustics 127 (2017) 63–73 73
Noise affects human health, causing annoyance, sleep disturbance and increasing the risk of cardiovascular disease. The quantification of noise impacts highlights it as a public health problem for which road traffic is mainly responsible. Life cycle assessment (LCA) is a technique to assess the environmental impacts of a product, a service or a process. Despite taking into account many environmental problems, the impact of noise on human health is not yet properly taken into account in LCA. The aim of this PhD thesis is to integrate the impact of traffic noise on human health in the LCA framework.
The scientific elements of acoustics and epidemiology that allow this integration are presented. An analysis of the existing methods is conducted by applying them to a case study. This helps to understand the advantages and drawbacks of the different approaches while comparing the results they provide. A method to integrate the impact of road traffic noise on human health in the LCA framework is then proposed. The method is based on noise prediction software and data made available by the Directive 2002/49/EC. This makes it possible to establish, with great precision, characterisation factors (CFs) connecting elementary flows of the LCA inventory with an impact on human health.
The method is then applied to a sample of small geographic areas selected in the region surrounding the city of Lyon (France). The application of the method and the analysis of the results provides a multitude of information regarding the potential existence of a typology for spatial differentiation, the best form for the collection of noise information at the LCA inventory level, the spatial variability of the CFs and the uncertainties that may be associated with them. The CFs obtained show that integrating the impact of noise into LCA could double the impact of road transport on human health. This PhD thesis also identifies further potential research topics. Similar work needs to be done for other transport modes (mainly trains and airplanes) to allow for a fair comparison of different transport modes in LCA studies. Repeating this method in other geographical areas with other acoustic emission and propagation models and/or other noise prediction software would also help the generalisation of this work and the assessment of possible sources of uncertainties.
Résumé
Le bruit affecte la santé humaine, provoquant de la gêne, des troubles du sommeil et augmentant le risque de crise cardiaque. Les quantifications de l’impact du bruit montrent que c’est un problème de santé publique et que le trafic routier en est majoritairement responsable. L’analyse du cycle de vie (ACV) est une méthode d’évaluation globale des impacts environnementaux d’un produit, d’un service ou d’un processus. Malgré la prise en compte de nombreux problèmes environnementaux, l’impact du bruit sur la santé humaine n’est pas encore correctement pris en compte dans l’ACV. L’objet de ce doctorat est d’intégrer dans l’ACV l’impact du bruit du trafic routier sur la santé humaine.
Les différents éléments d’acoustique et d’épidémiologie qui permettent cette intégration sont présentés. Une analyse des méthodes existantes est conduite en les appliquant à un cas d’étude. Cela permet de comprendre les avantages et inconvénients des différentes approches tout en comparant les résultats qu’elles fournissent. Une méthode pour intégrer l’impact du bruit du trafic routier sur la santé humaine dans l’ACV est ensuite proposée. Cette méthode repose sur les logiciels de prédiction acoustique et les données rendues disponibles par la directive 2002/49/CE. Elle permet d’établir, avec une grande précision, des facteurs de caractérisations (CFs) reliant des flux élémentaires de l’inventaire ACV à un impact sur la santé humaine.
La méthode est ensuite appliquée sur un échantillon de petites zones géographiques sélectionnées dans la région lyonnaise. L’application de la méthode et l’analyse des résultats apportent de nombreux enseignements sur l’existence potentielle d’une typologie pour la différentiation géographique, la meilleure forme pour la collecte d’information sur le bruit au niveau de l’inventaire ACV, la variabilité spatiale des CFs ou encore l’incertitude qui peut leur être associée. Les CFs obtenus montrent que l’intégration de l’impact du bruit en ACV pourrait doubler l’impact du transport routier sur la santé humaine. Ce doctorat identifie également des pistes de recherche. Des travaux similaires doivent être menés pour les autres moyens de transport (principalement trains et avions) pour permettre une comparaison équitable des études ACV les impliquant. Répéter la méthode dans d’autres zones géographiques, avec d’autres modèles d’émission et de propagation acoustique et/ou d’autres logiciels de propagation acoustique apporterait également des éléments intéressants.